Computer science internships

Supervisors: Prof Alan Fekete, A/Prof Uwe Roehm (collaboration with Prof David James in CPC)

Eligibility: 

Essential:

    -    Knowledge of SQL; experience installing software, administration and tuning, and dealing            with dependencies etc.

Desirable:

    -    Experience with document databases and/or graph databases; experience measur            performance of systems; some knowledge of bio-datasets.

Project Description:

A research project in the Charles Perkins Centre is building a data management system (using a relational approach with PostgreSQL) that allows integration of information from multiple bio-datasets, such as genome sequences, proteome sequences, and multiple experiments on biological samples. Research in Computer Science is exploring alternative data models and platforms. This project will be done with these researchers, aiming to provide a common workload to compare the performance of the alternatives.

Requirement to be on campus: No

Supervisors: Prof Alan Fekete, A/Prof Uwe Roehm

Eligibility:

• Essential: Knowledge of SQL; experience installing software, administration and tuning, and dealing with dependencies etc.

Project Description:

While the pure relational model keeps all data in simple fields (string, integer, etc), many modern dbms provide structured field types such as list, array, document. This project will explore the performance implications of keeping data in a structured field, comparing times for querying one component of the structure as the length and complexity of the structure varies. We will begin with PostgreSQL, if time or people permit, we will also look at MongoDB.

Requirement to be on campus: No

Supervisor: A/Prof Anusha Withana, A/Prof Craig Jin, Yihao Dong

Eligibility Criteria: 

    -    Basic knowledge of human-computer interaction, signal processing, or sensory             augmentation (preferred but not required).

    -    Familiarity with Unity, Python, MATLAB, C# or C++ (a plus).

    -    Strong interest in accessibility and assistive technology.

Project Description:

People with vision impairment (PVI), including those with low vision or blindness, face challenges in understanding their surroundings. Traditionally, they rely on white canes or guide dogs for navigation. However, limited research has explored how tactile cues, such as vibrotactile or electro-tactile feedback, can help PVI perceive their environment from a distance. For example, providing hierarchical spatial information through tactile feedback remains underexplored.

This project aims to explore the feasibility of using tactile cues to convey environmental information. We will start with simple geometric cues and gradually investigate more complex spatial details. By identifying effective tactile feedback strategies, this research seeks to lay the groundwork for future innovations in assistive technology for PVI.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisor: Dr Anusha Withana

Eligibility: 

    -    Programming skill in Python

    -    Familiarity with LLMs and prompt engineering

     -    Experience with data analysis or experimental evaluation

       -  Interest in AI for education, human–AI interaction

Project Description:

Students are the primary user of LLM, as it can provide explanations and feedback quickly. With the growing adoption of tutor and long horizon agent, the increasing reliance raises concerns about misleading or overly direct answers that limit deeper comprehension.  Different LLM personas and styles may significantly influence meaningfulness and effectiveness of learning. There is currently no standard benchmark for evaluating LLM personas, most existing benchmarks focus on task accuracy or reasoning ability.

To address this gap, to develop a benchmark dataset which consists of tutoring tasks and an evaluation metrics is possible. Systems can interact with these tasks, and their performance will be evaluated using metrics.

By providing an open-source benchmark that allows researchers to evaluate systems and models, this work also helps to enhance the trustworthiness of LLM and tutor agents for student, reduce negative learning outcomes, and support the design of more effective AI tutors.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisor: A/Prof Anusha Withana, Prof Alistair McEwan, Praneeth Perera

 Eligibility:

    -    Programming skill in Python and C# (Unity development)

    -    Basic electronics knowledge

Project Description:

Virtual reality (VR) is getting so popular. But VR does not let us reaching out to touch a virtual object in VR and actually feel its shape, weight, or even its movement against your skin. This project aims to bridge the gap between seeing a digital world and truly physically experiencing it. We are developing a neural communication system that directly stimulate nerve fibers to generate sensations like the buzz of a button or the resistance of a heavy object directly to your fingertips and muscles.  We envision "feelable" virtual environments where the boundary between the digital and the physical starts to disappear. It’s a mix of creative world-building, smart hardware, and human-centered design. As an intern, you will integrate the system with virtual environments and create new haptic experiences.

Requirement to be on campus: Yes *dependent on government’s health advice

Supervisor: Prof Athman Bouguettaya

Eligibility:

    -    Good Programming Skills; Data Handling and Management Skills

    -    WAM=>85

Project Description:

The Internet of Things (IoT) is transforming our world into a seamlessly connected ecosystem of smart devices. IoT devices collaborate by sharing resources, such as energy and computation, through IoT service crowdsourcing. However, such services can be exploited; for example, a consumer may misuse a shared Wi-Fi hotspot to launch a Denial-of-Service (DoS) attack. Trust management frameworks have been introduced to address these risks by establishing and maintaining trust among entities. While these frameworks provide robust trust assessment methods, a critical challenge remains: ensuring the integrity of trust information. Since distributed entities are responsible for storing and managing trust data, they may tamper with it, making the system vulnerable to internal attacks. The rise of AI tools, such as ChatGPT, has further lowered the barrier for adversaries, enabling even low-skilled actors to manipulate trust data with alarming sophistication. To address this threat, this project explores AI-based tampering detection techniques and environmental visualization of the IoT crowdsourced environment.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisor: Prof Athman Bouguettaya

Eligibility: 

    -    Good Programming Skills; Data Handling and Management Skills

    -.   Experience in Drone Flight Simulators will be a plus

    -    WAM=>85

Project Description:

A continuous expansion of urban areas is leading to an increased demand for instant deliveries from warehouses to customers' doorsteps. Unmanned Aerial Vehicles (UAVs) or drones have the potential to serve customers with timely and cost-effective deliveries. Drones usually operate in a skyway network, which is an interconnected set of nodes. The nodes are building rooftops that serve as recharging stations or delivery destinations for drones. Drones may recharge at nodes for long distance deliveries as they are constrained by limited battery capacity. These nodes are connected through skyway segments, which multiple drones may share at the same time to transit between nodes. However, managing drone traffic in congested urban environments presents significant challenges. The risk of aerodynamic interference among multiple drones operating in shared skyway segments may impact drone delivery efficiency and safety. The focus of this project is to develop an AI-driven stabilization system for drones to counteract the effects of interference caused by peer drones or environmental factors. We leverage machine learning techniques and real-time sensor data to identify aerodynamic disturbances and generate corrective control actions. The goal is to ensure safe and efficient operation of drones in a shared skyway network.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisor: A/Prof Chang Xu

Eligibility: 

    -    Strong programming skills in Python and experience with PyTorch

    -    Experience with Deep Learning

    -    Familiar with command line tools, version control and experiment tracking tool

    -    Good written and spoken English communication skills

Project Description:

This project aims to build an AI-driven pipeline for mRNA sequence design to improve protein expression and functional duration in cells. The project focuses on learning sequence-to-function relationships from public data and in-house experiments and using these models to propose optimized mRNA constructs under practical biological constraints. You will work with sequence modelling methods (e.g., transformers and lightweight predictors), feature extraction (secondary structure, motif constraints), and multi-objective optimization to generate candidate 5′UTR and coding sequences for a target protein. The work includes building reproducible data and evaluation pipelines, designing candidate libraries, and supporting iterative experimental validation to continuously improve model quality. Students will gain hands-on experience at the intersection of deep learning, bioinformatics, and mRNA therapeutics.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisor:  Dr Clément Canonne

Eligibility: Having taken COMP3027 or COMP3927 (or equivalent) with a DI or HD, solid background in discrete mathematics.

Project Description:

This project will focus on understanding the power, limitations, and applications of quantum pseudodeterministic algorithms, as introduced by Aaronson, Gur, and Li (https://arxiv.org/abs/2602.17647v1): that is, quantum algorithms which, in spite of the inherent randomness of quantum computation, can be made to (nearly) always output the same answer.

Requirement to be on campus: No

Supervisor:  Dr Clément Canonne

Eligibility: Having taken COMP3027 or COMP3927 (or equivalent) with a DI or HD, solid background in discrete mathematics.

Project Description:

This project will focus on understanding and developing new quantum algorithms in the interactive setting, where a (limited) algorithm, Arthur, interacts with another (all-powerful, but unreliable) algorithm, Merlin, to solve the task at hand. Here, we will specifically consider the setting where (1) both Arthur and Merlin are quantum algorithms, and (2) the task to be solved is a data analysis (learning or testing) task, on an underlying large dataset or probability distribution.

Requirement to be on campus: No

Supervisor: Prof Joseph G. Davis

Eligibility: WAM>75 and Undergraduate candidates must have already completed at least 96 credit points towards their undergraduate degree at the time of application.

Project Description:

Behavioural biometrics are crucial for continuous authentication, particularly as zero-trust architectures gain prominence. Systems can verify user identity in real-time by analysing multimodal data streams – such as keystroke dynamics, mouse movements, gait, and touch patterns. This continuous verification enhances security by detecting anomalies and preventing unauthorised access. However, real-world data is often noisy, incomplete, and subject to environmental variability, posing significant challenges for reliable classification and detection.

To this end, this project offers two potential directions. The first focuses on developing and evaluating JEPA-based models by Meta to learn robust representations from behavioural biometrics data, addressing inconsistencies and improving predictive accuracy.

The second is to explore whether the in-context learning abilities of sensor foundation models can be used for reliable continuous user authentication. This direction is based on the recent work from Google.

Requirement to be on campus: No

Supervisor: Dr Lijun Chang

Eligibility: Good algorithm design and C (or C++) programming skills.

Project Description:

We are nowadays facing a tremendous amount of large-scale social networks with millions or billions of edges. Thus, there is a need of designing efficient algorithms for processing large-scale graphs. In this project, our aim is to design efficient algorithms to speed up graph processing on ever-growing large graph datasets. Example problems that we will be investigating include efficiently finding a large near-clique over a large sparse graph, efficiently enumerating all maximal subgraphs that satisfy a certain condition.

Requirement to be on campus: No

Supervisor: Dr Liyi Zhou

Eligibility: WAM>75  and Undergraduate candidates must have already completed at least 96 credit points towards their undergraduate degree at the time of application.

Project Description:

Cybercrime is growing quickly and causes large financial losses. In Australia alone, the ACSC reported more than 87,400 cybercrime incidents in 2023 to 2024, with major costs for individuals and small businesses. Current software security tools depend heavily on predefined patterns and expert knowledge. Experts must define vulnerability patterns, analyze results, and confirm issues manually. This process is expensive, slow, and difficult to scale as software systems continue to grow.

Recent work has explored using large language models to detect vulnerabilities. These approaches mainly rely on pattern recognition, fine tuning, or prompt engineering. While useful, they remain passive, require human supervision, and often produce false positives or miss real vulnerabilities.

New AI models with stronger reasoning abilities suggest a different path. This project aims to build an end to end AI security agent that can analyze software, discover vulnerabilities, generate attack strategies, and verify them automatically. Blockchain systems provide a suitable testbed because they contain many real world exploits in a structured execution environment.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisor: Dr Mohammad Polash

Eligibility: 

    -    Students with an interest in NLP, machine learning, education technology, and Python.

    -    Experience with text analytics, embeddings, or dashboard development would be helpful.

    -    Must be willing to write a research article about this.

Project Description:

This project aims to build an AI-based educational analytics system that helps instructors identify the concepts students find most confusing in a UoS. The system will analyse student questions, forum posts, assignment errors, quiz responses, and feedback comments to detect common misconceptions and recurring learning bottlenecks. Likely methods include clustering, topic modelling, semantic embeddings, misconception classification, and difficulty trend analysis. Results will be presented in a teacher-facing dashboard that shows confusing topics, repeated questions, student-topic patterns, and recommended interventions.

Requirement to be on campus: No

Supervisors: Dr Mohammad Polash

Eligibility:

    -    Students with an interest in NLP, information retrieval, machine learning, and Python.

    -    Familiarity with transformers, embeddings, or search systems is desirable.

    -    Must be willing to write a research article about this.

Project Description:

This project proposes an AI-driven literature discovery system that recommends semantically similar research papers based on content, methodology, and citation relationships. Using transformer-based scientific text embeddings, citation graph analysis, topic modelling, and similarity search, the system will go beyond keyword matching to find conceptually related work across disciplines. A user may upload a paper or enter a query and receive related publications, topic clusters, citation insights, and author recommendations. The project can improve research efficiency, support interdisciplinary discovery, and contribute to information retrieval research. Extensions may include benchmark identification, trend analysis, and the discovery of research gaps.

Requirement to be on campus: No

Supervisor: Dr Mohammad Polash

Eligibility:

         -     Ability to review current literature on this topic

          -    Have the skillset to implement such a too, i.e. experience working with LLM API

          -    Willing to write a research article about this.     

Project Description:

With the rapid rise of generative AI, students increasingly rely on AI-generated code to solve programming tasks. While this accelerates problem-solving, it often undermines their confidence in writing code independently. This project addresses this challenge by developing an educational tool that leverages GenAI to strengthen students’ programming skills and computational thinking.

The tool generates intentionally buggy code based on a given problem specification and prompts students to critically analyze it. Learners will be tasked with identifying logical, syntactical, or structural errors, encouraging a deeper understanding of programming concepts. Additionally, the tool guides students to translate problem requirements into step-by-step computational logic, reinforcing their ability to design solutions before coding.

By shifting the focus from passive code consumption to active problem-solving and debugging, the project aims to build students’ confidence, autonomy, and resilience in programming, equipping them with skills crucial for both academic and professional success.

Requirement to be on campus: No

Supervisor: Dr Muhammad Sajjad Akbar and Dr Mohammad Polash

Eligibility: Machine Learning, it would be good to have security background but not compulsory.

Project Description:

Digital accessibility education often focuses on physical disabilities, while the experiences of neurodivergent users—including individuals with ADHD, autism, or cognitive processing differences receive less attention. Many interface designs unintentionally create barriers such as information overload, complex navigation structures, or distracting visual elements.

This project explores the use of serious games as a learning tool to help students understand accessibility challenges experienced by neurodivergent users. The proposed platform will simulate interface environments where students must complete tasks under conditions that replicate cognitive overload, sensory distractions, or time pressure. These experiences will highlight the importance of clear structure, reduced cognitive load, and accessible interface design.

The game will include interactive design challenges where students modify interface elements to improve accessibility. For example, they may redesign menus, simplify instructions, or adjust visual layouts to support different cognitive needs.

By transforming accessibility learning into a hands-on game-based experience, the project aims to improve student understanding of inclusive design principles.

The research will investigate how game-based simulations can enhance awareness, empathy, and accessibility design skills among computing students.

Requirement to be on campus:  No

Supervisor: Dr Muhammad Sajjad Akbar and Dr Mohammad Polash

Eligibility: Machine Learning, it would be good to have security background but not compulsory.

Project Description:

Accessibility is an essential topic in computing education, yet many students struggle to understand the real-world challenges faced by users with disabilities. Traditional teaching approaches often rely on lectures or written guidelines, which may not fully convey the lived experiences of users with accessibility needs.

This project proposes the development of a Virtual Reality (VR)-based gamified learning environment to support accessibility education in large cohort units such as INFO5990 and INFO6007. The system will simulate interactive scenarios where students experience digital interfaces from the perspective of users with visual, auditory, or motor impairments. Through immersive gameplay, students will complete tasks such as navigating inaccessible websites, designing improved interfaces, and solving accessibility challenges.

Gamification elements such as levels, rewards, and scenario-based missions will encourage exploration and experimentation. The VR environment will help students understand accessibility barriers and the importance of inclusive design.

The research will evaluate whether immersive VR experiences improve student empathy, accessibility awareness, and practical design skills compared to traditional learning approaches.

Requirement to be on campus: No

Supervisors: Dr Muhammad Sajjad Akbar and Dr Mohammad Polash

Eligibility: Machine Learning, it would be good to have security background but not compulsory.

Project Description:

Large cohort courses such as INFO5990 and INFO6007 at the University of Sydney often involve hundreds of students attending tutorials. While these sessions are designed to promote discussion and collaborative learning, student engagement can be inconsistent. Many students remain passive observers rather than active participants due to time constraints, large class sizes, or hesitation to speak in group settings. As a result, opportunities for formative learning, peer discussion, and immediate feedback are reduced.

This project proposes the design and development of an AI-based gamification platform to increase student participation during tutorials. The tool will transform tutorial activities into structured game-based challenges where students earn points, badges, and progression levels by solving short problems, answering concept questions, or participating in group discussions. AI techniques will dynamically generate quiz questions, adapt difficulty levels based on student responses, and provide immediate explanations to support learning.

The platform will integrate with common teaching tools and allow tutors to run live tutorial challenges, track participation, and identify students who may require additional support. Gamification elements such as leaderboards, cooperative tasks, and time-limited missions will encourage active involvement without increasing tutor workload.

The project will evaluate whether AI-driven gamification improves student engagement, participation rates, and learning outcomes in large tutorial environments.

Requirement to be on campus: No

Supervisors: Dr Muhammad Sajjad Akbar and Dr Mohammad Polash

Eligibility: Machine Learning, it would be good to have security background but not compulsory.

Project Description:

In large units, instructors often manage hundreds of students and multiple assessments, creating significant challenges in providing timely and meaningful feedback. Tutors may require considerable time to review assignments, and delays in feedback reduce its effectiveness for learning improvement. Students also frequently request clarification on grading decisions, further increasing tutor workload.

This project proposes the development of an AI-assisted feedback and marking support tool designed specifically for large cohort environments. The tool will use natural language processing and semantic similarity models to compare student responses with rubric criteria and sample solutions. Instead of fully automated grading, the system will assist tutors by highlighting key concepts, identifying missing reasoning, and suggesting feedback aligned with the rubric.

The tool will also generate instant formative feedback for students immediately after submission, helping them understand areas of improvement before final grading. Tutors will review AI-generated suggestions and make final decisions, ensuring academic judgment remains central to the evaluation process.

The system will include features such as rubric mapping, automated comment generation, and batch analysis of submissions, enabling tutors to maintain consistency while reducing marking time.

The project will investigate whether AI-assisted marking can improve feedback quality, reduce grading workload, and enhance learning outcomes in large cohort courses.

Requirement to be on campus: No

Supervisor: Dr Muhammad Sajjad Akbar and Dr Mohammad Polash

Eligibility: Machine Learning, it would be good to have security background but not compulsory.

Project Description:

The increasing availability of AI text generators has significantly changed the nature of open assessments in higher education. In large units such as INFO5990 and INFO6007, where many assignments are completed outside the classroom, students may rely on generative AI to produce answers without engaging deeply with the learning material. This creates a risk of “illusion of learning,” where students believe they understand concepts simply because they can generate answers.

This project proposes the development of an AI-resilient assessment design tool to support Lane 2 assessments, which focus on student-centred, open-ended learning activities. The proposed system will analyse assessment questions and evaluate their AI-solvability risk using language models and semantic analysis. Based on this analysis, the tool will recommend redesign strategies that encourage higher-order thinking such as reflection, personal context, iterative reasoning, and applied problem solving.

The tool will also generate student-centric prompts, scaffolded questions, and metacognitive reflection activities that reduce reliance on AI-generated answers. Tutors and instructors will receive recommendations for improving assessment structure while maintaining usability and fairness.

The project will evaluate how AI-informed assessment redesign can support authentic learning, critical thinking, and student engagement in large cohort environments.

Requirement to be on campus: No

Supervisors: Dr Rahul Gopinath, Prof Alan Fekete, A/Prof Uwe Roehm

Eligibility: Some experience with automated testing; also experience with Go language is desirable.

Project Description:

The Ethereum blockchain relies on a database layer to manage persistent state, yet this layer remains largely untested by automated methods. This project investigates its robustness by applying fuzzing — a technique that generates diverse, unexpected inputs to expose failures and inconsistencies in software systems.

Building on prior work by the supervisors and vacation research students — who applied blackbox grammar inference techniques (L* and RPNI) to characterise and probe the Ethereum database layer — this project develops more targeted grammar refinement strategies using advanced inference methods, and fuzzing guided by the resulting specialised grammars.

Requirement to be on campus: No

Supervisors: Dr Rahul Gopinath,  Dr Hong Jin Kang

Eligibility: You will work with the supervisor directly for this project. You should be a fast learner. Excellent skills in programming, using LLMs, as well as the ability to work independently are required.

Project Description:

Software testing research has been built on the "competent programmer hypothesis" - the assumption that most bugs in code are relatively simple mistakes made by otherwise skilled programmers, like using the wrong comparison operator or being off-by-one in a loop condition. This assumption underlies mutation testing, a powerful technique that evaluates test suites by introducing small, artificial faults into programs to see if the tests can catch them. However, as Large Language Models (LLMs) like GPT-4 and Claude increasingly write code, we face a fundamental question: do these AI systems make the same kinds of mistakes that human programmers do?

This research investigates whether the error patterns of LLM-generated code follow the same distributions as human errors. By having LLMs describe functions in natural language and then re-implement them, we can study how their error models differ from humans and determine whether our existing testing methodologies remain effective in a world where much of our code may be written by artificial intelligence rather than human hands.

This project may be extended to an Honours thesis with the supervisor, and if completed successfully could be the basis of a publication at a top SE venue.

(In your application, please indicate if you are interested in a Honours application with the supervisor).

Requirement to be on campus:  No

Supervisor: Rahul Gopinath,  Dr Hong Jin Kang

Eligibility: Should know C and Python well.

Project Description:

For fuzzing commands (find, gcc, etc.) to be effective, we need to not only exercise all of the options possible, but also ensure that option combinations are exercised. This is impossible to do with the current state of the art because typical fuzzers can waste time generating invalid option combinations. An alternative is to derive a grammar of the input options, leveraging example options from documentation, and using techniques such as RPNI, TTT etc.

In this project, you will implement a grammar inference mechanism built on the RPNI grammar inference algorithm, and use it to infer option grammar of commands such as GNU find, GCC, CLANG etc. and use it to fuzz.

This project may be extended to an Honours thesis with the supervisor, and if completed successfully could be the basis of a publication at a top SE venue.

(In your application, please indicate if you are interested in a Honours application with the supervisor).

Requirement to be on campus: No.

Supervisor: Dr Rahul Gopinath, Dr Hong Jin Kang

Eligibility: The student should have prior experience in machine learning and distributed computing. The project is well-suited for students who are motivated to take Research Pathway Project, Honours Research Project, MPhil, or PhD.

Project Description:

During data engineering, many records fail validation due to trivial errors (e.g. invalid delimiters in date format). This can cause loss of accuracy due to lost records. Hence,

Data Repair or Data Imputation is an important part of data engineering and data science. Current methods for data imputation focus on numerical data and hence, use numerical data. However, the question is, can we compute a closest match to corrupted records, and provide any empirical assurance for the computed match, that it is the right repair, and is closest to data seen? While naive data repair is possible, it can be computationally expensive and may lead to data loss.

In this project you will implement a data imputation system using inferred grammar using an active learning strategy such as L* or TTT. Then, it will be compared against competitors to determine the effectiveness of data thus imputed.

This project may be extended to an Honours thesis with the supervisor, and if completed successfully could be the basis of a publication at a top SE venue.

(In your application, please indicate if you are interested in a Honours application with the supervisor).

Requirement to be on campus: No

Supervisor:  Dr Rahul Gopinath, Dr Hong Jin Kang

Eligibility: Should know C and Python well.

Project Description:

During fuzzing, we often find crashes resulting from inputs that are very large (GBs in size). To understand why a crash happened, it is important to minimize that input, so that the core issue is clear. Delta debugging is the classical algorithm for input minimization. However, its effectiveness in minimizing structured inputs is limited. While input grammars can help, they are often not available, or is inapplicable (e.g. in non-standard inputs in compiler testing that may not be parseable). A way forward is to use the delta debugging as a mechanism to generate example inputs, and use these inputs to derive an input grammar for the crash in question; then minimize the input using the inferred grammar.

In this project, you will implement an input minimization algorithm using inferred grammar, and compare that against competitors for input minimization.

This project may be extended to an Honours thesis with the supervisor, and if completed successfully could be the basis of a publication at a top SE venue.

(In your application, please indicate if you are interested in a Honours application with the supervisor).

Requirement to be on campus: No

Supervisors: Dr Sasha Rubin

Eligibility: WAM>75 and Undergraduate candidates must have already completed at least 96 credit points towards their undergraduate degree at the time of application.

Project Description:

Planning is part of the symbolic/logic approach to AI which involves finding a finite- state program that tells an agent what to do in every state. The states and possible actions are described declaratively. See, e.g., https://research.ibm.com/projects/ai-planning

One approach to building a planner is to reduce the given planning problem to the satisfiability problem, and implement this as a reduction to SAT solvers or in declarative programming languages such as ASP (https://potassco.org/)

There are a few possible topics for this project, depending on the interest and skill of the student.

E.g., Build a declarative planner for modern (decision-theoretic) solutions such as non-dominated solutions.

E.g., Build a declarative planner that solves “lifted planning” for standard solutions (such as strong or strong-cyclic) by reducing to automated theorem proving.

All of the work will have some mathematical component, and may include a coding component depending on the topic. This project will suit a student who achieved an HD in a course on Theoretical Computer Science.

Requirement to be on campus: No

Supervisors: Prof Seokhee Hong, Dr Amyra Meidiana

Eligibility: 

    -    Data Structure and Algorithms

    -    Programming (Java, C++, Python, Javascript)

Project Description:

Technological advances have increased data volumes in the last few years, and now we are experiencing a “data deluge” in which data is produced much faster than it can be understood by humans. These big complex data sets have grown in importance due to factors such as international terrorism, the success of genomics, increasingly complex software systems, and widespread fraud on stock markets.

Visualisation is a powerful tool to compute good geometric representation of abstract data to support analysts to find insights and patterns in big complex data sets.

This project aims to design, implement and evaluate new visualisation algorithms for scalable and faithful visualisation of big complex data, to enable humans to find ground truth structure in big complex data sets, such as social networks and biological networks.

These new visualisation methods are in high demand by industry for the next generation visual analytic tools.

Requirement to be on campus: No

Supervisors: Sri AravindaKrishnan Thyagarajan

Eligibility: WAM>85

Project Description:

In the data-driven world, information drives industry, commerce and scientific development. Quite often information is outsourced and accessing them is as sensitive as the content in the information itself. Consider a privacy-preserving internet search. How often do we wish your search is not monitored by the search engine, but we should get the correct answer. This problem is tackled by Private information retrieval or PIR. All major tech giants rely on PIR at critical parts of their infrastructure.

In this project, we will study how to secure PIR in the post quantum world where attackers may have quantum computing capabilities. We will look at existing post-quantum PIR schemes and improve them along different axes (security, privacy and efficiency).

This project will be apt for you if you are interested in working with modern cryptography, and if you have a good grasp of linear algebra, algorithms, complexity and probability.

Requirement to be on campus: No

Supervisors:  Dr Wei Bao

Eligibility: The student should have prior experience in machine learning and distributed computing. The project is well-suited for students who are motivated to take Research Pathway Project, Honours Research Project, MPhil, or PhD.

Project Description:

Federated learning (FL) enables collaborative model training without centralising data, but it raises critical concerns about the “right to be forgotten.” Federated unlearning (FU) aims to remove the impact of specific data from trained models. However, existing methods often overlook adversarial environments, where malicious data removal requests could compromise the model’s integrity and reliability. This project will explore novel methods to make FU algorithms more robust and efficient against adversarial behaviours, while ensuring model accuracy and fairness.

The student will design algorithms that can identify and mitigate the risks of malicious removal requests. The work will involve implementing these algorithms, and conducting experiments on benchmark datasets.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisors: Dr Yunke Wang

Eligibility: Applicants should have practical experience in at least one of the following:

    -    CAD mechanical design (e.g., SolidWorks)

    -    Embedded systems

    -    Hands-on experience with physical hardware is highly desirable.

Project Description:

This project involves the design and prototyping of a mobile robotic platform to support ongoing experimental research. The student will be responsible for mechanical structure design using CAD tools (e.g., SolidWorks), component selection, structural optimisation, and fabrication-ready modelling, followed by full system assembly.

The project also includes embedded system development for motor control, sensor interfacing, and low-level firmware implementation. The student will contribute to hardware–software integration, system debugging, and iterative performance testing to ensure mechanical robustness, control stability, and operational reliability.

Requirement to be on campus: Yes *dependent on government’s health advice

Supervisors: Dr Zhanna Sarsenbayeva, A/Prof Anusha Withana

Eligibility: You will work with the supervisor and a PhD student, and we expect you are a fast learner. Familiarity with point cloud processing required. Blender proficiency highly desirable.

Project Description:

This project focuses on the development of an AI voice health coach embedded within a mobile application to promote sustained physical activity. The system tracks users’ daily

activity data (e.g., step count) against self-defined goals and delivers personalized motivational support.  The voice coach will engage users in dynamic motivational

conversations using distinct communication styles, such as empathetic and directive approaches. An integrated NaturalLanguage Processing (NLP) module will enable the system tounderstand user responses, interpret motivational states, and adapt conversations accordingly. The application will also administer brief daily questionnaires and reminders tomonitor engagement and behavioural progress.  The platform will be built with scalability in mind, allowingseamless integration with wearable devices for real-time activity tracking. Overall, the project combines conversational AI, behavioural science, and mobile health technologies tosupport long-term habit formation and user engagement.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisor: Prof Zhiyong Wang

Eligibility: 

    -    WAM >= 80

    -    Strong computer science background

    -    Good knowledge of bioinformatics

Project Description:

Recent breakthroughs in Artificial Intelligence (AI) have led phenomenal success in protein structure prediction, which presents immense opportunities to accelerate the innovations in structure based drug discovery. This project aims to develop novel generative AI methods by exploiting the structures of proteins for effective drug design. 

Requirement to be on campus: No

Supervisor: Prof Eduardo Velloso (in collaboration with the Brain and Mind Centre)

Eligibility: Experience with web and/or mobile development.

Project Description:

Financial stress is a growing contributor to psychological distress among young people, particularly in digital environments where spending and betting occur rapidly and with little reflection. This project explores how interactive technologies can help people develop self-regulation micro-skills that support healthier decision making and reduce financial stress.

Students will contribute to the design and evaluation of a digital mental-health prevention tool that uses personal spending data to provide reflective feedback and short interactive learning modules. The project combines human–computer interaction, digital health, and behavioural science.

Interns will assist with activities such as analysing existing digital tools, designing interface concepts, prototyping interactive features, and evaluating usability with users. The work will contribute to an early-stage research prototype being developed by a cross-faculty team in Engineering and Medicine.

The project provides experience in interaction design, digital health research, and interdisciplinary collaboration.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisor: Dr William Zhi

Eligibility: 

    -    Strong Python skills required. Prior exposure to computer vision,

    -    3D geometry, deep learning, or robotics is desirable. Experience with PyTorch, Open3D,            or related tools is helpful.

Project Description:

This project investigates how 3D foundation models can help robots perceive and use objects while they are being grasped. Thestudent will explore methods for estimating the geometry and poseof held objects from RGB or RGB-D observations, and examine how these representations support downstream tasks such as tool use, motion planning, or camera-robot calibration. The project maycombine recent ideas in foundation-model based pose estimation, multi-view geometric consistency, and uncertainty-aware scene representation. Using public datasets or simulation, the student will benchmark alternative pipelines and analyse which representations generalise best to novel objects, occlusion, and

viewpoint change. The project connects robot perception with manipulation and contributes to current research in embodied AI and 3D vision.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisor: Dr William Zhi

Eligibility: Background in Python programming required. Experience with machine learning, robotics, computer vision, or PyTorch preferred. Suitable for students in computer science, software engineering, robotics, or related quantitative disciplines.

Project Description:

This project explores how robots can learn manipulation tasks from hand-drawn sketches and other lightweight human inputs. The student will study sketch-guided task specification for tabletop manipulation, combining spatial sketch representations with modern vision-language or multimodal foundation models to infer task constraints, object relations, and target motions. Using simulation or public robotic datasets, the project will benchmark

methods for translating sketches into executable manipulation plans or policies, and evaluate robustness under ambiguous goals, clutter, and unseen scene layouts. The project sits at the intersection of robot learning, human-robot interaction, and embodied AI, and will contribute toward more intuitive ways for non-experts to teach robots new skills.

Requirement to be on campus: Yes *dependent on government’s health advice.

Supervisor: Prof Judy Kay

Eligibility:  WAM>75 and Undergraduate candidates must have already completed at least 96 credit points towards their undergraduate degree at the time of application.

Project Description:

Investigative interviewing involves obtaining information from people about events to support decision‑making. Professionals in forensic, health, education, and business settings frequently conduct such interviews but often receive limited training, which can lead to poor interviews and harmful decisions.

AI chatbots offer a promising solution for scalable, simulation‑based interview training. They can provide consistent appropriate responses and immediate feedback.

At the University of Sydney, academic progression advisors receive no interview training yet regularly interview at-risk students about their academic and personal challenges. This project will develop a chatbot training system that enables advisors to rehearse advising sessions with realistic, multi‑turn dialogue. The system will include a dialogue model, a learner model to track skill development, and an intuitive interface. This project could explore a variety of dialogue approaches, such as multi‑turn question‑answering agent that queries a structured knowledge base or performs conversational machine reading of mock case facts.

Requirement to be on campus: No

Supervisor: Dr Clément Canonne, Dr Kanchana Thilakarathna

Eligibility: Strong coding skills, experience with LLMs, background in algorithms (COMP3027 or equivalent).

Project Description:

This project will explore several aspects of training and fine-tuning AI and models on synthetic data, focusing on synthetic data generated using various privacy-prerserving methods.

This exploratory project will consider three main axes: (1) (quality) how good this approach is, in terms of the resulting quality and performance of the models; (2) (exploration power) how good of a proxy this privacy-preserving training can be to understand the ultimate performance of models training on real data; and (3) (feasibility) what are the computational aspects, i.e., does this lead to an acceptable, or prohibitive, overhead of the synthetic data generation in terms of computational resources during the training phase, and its impact on the duration of that training to achieve good accuracy.

Requirement to be on campus: No

Supervisors: Dr Clement Canonne and Dr Sri Aravinda Krishnan Thyagarajan

Eligibility: Having taken COMP3027 or COMP3927 (or equivalent) with a DI or HD, solid background in discrete mathematics.

Project Description:
In classical computers we know we can execute a program multiple times. Just repeat the run with different inputs. But in the quantum computing environment, we can do something fascinating. This project will explore recent work on quantum one-time programs (Gupte, Liu, Raizes, Roberts, Vaikuntanathan, 2025: https://arxiv.org/abs/2411.01876): i.e., quantum algorithms that the owner Alice can give to a user Bob so that Bob can execute the algorithm only once. The internship will focus on understand the construction, limitations, and proof of correctness of these one-time programs, and potential new applications for cryptography. At the end of the internship, you would have grasped the basics if not more of quantum computing and quantum cryptography.

Requirement to be on campus: No

Supervisor: Dr Katy Gero

Eligibility:

Experience with qualitative methods (e.g. interviewing) and basic web development skills.

Project Description:

Due to the rapid introduction and adoption of large language models (LLMs) in universities, there is a lack of understanding of their impact on writing education. Much work has reported positive effects of generative AI on writing. However, there is also evidence that the use of generative AI may accumulate 'cognitive debt' and decrease or change the nature of critical thinking. This project investigates how teachers and students can use generative AI to improve learning outcomes in writing while safeguarding students' critical thinking. The research will involve: 1) exploring current usage of generative AI in university writing education; 2) prototyping and evaluating novel, learning-focused interventions.

Requirement to be on campus: Yes*dependent on government’s health advice