Modern AI agents —chatbots, autonomous planners, multimodal assistants — are bottlenecked by what they remember and how they retrieve. This joint PhD position, supervised across ANU and A*STAR Singapore, places you at the center of building the memory and retrieval infrastructure that the next generation of AI agents will run on.
Research Direction
Embedding unstructured data (text, images, audio) into high-dimensional vectors has become the default representation for AI workloads. The system that stores, indexes, and serves these vectors is the vector database. As AI agents move from single-shot chat to long-running, tool-using, multi-modal systems, there are challenges emerging. For example,
• Retrieval efficiency. Production ANN search must respond at millisecond latency over million-to-billion-scale vectors. How do we co-design retrieval indexes with the GPUs, memory hierarchies, and energy budgets they will actually run on?
• Adaptive memory. AI agent memory should grow, prune, and reorganize over time. How do we keep retrieval correct and fast under dynamic, streaming updates?
The joint project spans the boundary of database systems, machine learning, and AI infrastructure. You will publish at top-tier venues (VLDB, SIGMOD, ICDE, NeurIPS, ICML, etc.) and have access to two complementary research ecosystems.
This position is under the primary supervision of Dr. Mengxuan Zhang, along with one researcher from ANU and one from A*STAR to form the supervisory panel.
Program Structure
You will be enrolled at ANU, and your degree will be conferred by ANU. The program spans across two countries:
• Phase 1 — ANU (Canberra, Australia): 1–2 years. Coursework, foundational research, and paper preparation.
• Phase 2 — A*STAR(Singapore): 1–2 years. Research attachment under the Singapore Research Attachment Program (SRAP).
• Phase 3 — ANU: return for thesis writing and defense.
Total program length: 3–4years. The split is flexible based on research progress; every student will spend a minimum of 1 year physically at A*STAR Singapore under SRAP.
Funding and Benefits
At ANU (initial and final phases)
• ANU PhD stipend: AUD39,069 per annum (Full-time base stipend rate 2026)
• Tuition fee waiver available
At A*STAR Singapore(research attachment)
• Monthly stipend of S$3,600
• Monthly housing subsidy of S$300
• One-time airfare grant of S$1,500
• One-time settling-in allowance of S$1,000
• One-time IT allowance of S$800
• Conference allowance upto S$1,000 (local) and S$4,000 (overseas)
• Medical insurance
Note: Terms of award may be reviewed periodically
Eligibility
• International student(non-Singaporean, non-Singapore PR) — required by SRAP
• Not a previous awardee ofthe A*STAR Research Attachment Programme (ARAP)
• Solid programming background — C++ preferred — and a strong foundation in algorithms and data structures
• Familiarity with machine learning / deep learning fundamentals
• Genuine interest in systems-level research at the intersection of databases and AI
What You Get from Working in My Lab
• Weekly one-on-one mentoring with hands-on research guidance
• Top-tier publications(VLDB, SIGMOD, ICDE)
• International research network across Australia, Singapore, the United States, Hong Kong, and beyond
• Two-country PhD experience — both Canberra’s research-focused setting and Singapore’s industry-dense ecosystem
• A clear pathway to academia or industry post-PhD
How to Apply
Please email Dr. MengxuanZhang at Mengxuan.Zhang@anu.edu.au with the following:
• Subject line: “PhD Application — Memory and Retrieval Infrastructure (Joint ANU / A*STAR)”
• Your CV
• Academic transcripts(undergraduate and any postgraduate)
• A short statement (one page or less) describing your research interests and why this joint program fits you
• Names and contact details of two academic referees
Strong candidates will be invited for a video interview. Following that, formal applications must be submitted through the ANU PhD admission system. Applications are reviewed on a rolling basis — we encourage early submission.