Agentic-ready planetary-scale distributed geodata structure
Data structures and a programming model for querying a planetary-scale graph of landscape entities - largely by AI agents.
- PhD
- MS (Research)
- Pre-doc / Research Fellow
Explore what SIT faculty and students are working on. Filter by thrust area or by status:
Projects that list who they are looking for are open to PhD, MS (Research), pre-doctoral (Research Fellow) or RA candidates; see projects open to students. Interested in one? Write to the faculty contact listed. If you do not yet meet admission criteria, consider the Research Fellows pathway.
Thrust Area
Status
29 projects
Data structures and a programming model for querying a planetary-scale graph of landscape entities - largely by AI agents.
Version control and recomputation standards for DAGs of datasets and algorithms that span distributed hosting and compute nodes, including GPU-scaled algorithms.
Reducing the energy cost of embedding and retrieving across structured, semi-structured and unstructured data.
Refer to the owncloud document.
Building on a Hindi adaptation of Melodic Intonation Therapy developed with AIIMS Delhi, evaluating LLM-based mobile apps for diagnosis and treatment.
ML, programming, mobile app development, LLMs, interest in medical applications
AIIMS Delhi

Generative-AI and GPU-accelerated simulation to automatically produce PCB and 3D engineering designs.
Analysing VPNs, Tor and mixnets, regional app security and privacy, and building defences against online scams.
Applied networking, coding and scripting

Improving latency, throughput and cost of ML inference serving so that the shift from training to inference does not cost the planet.
Scaling model checkers and proof tools to gray and metastable failures, with LLM-based proof agents to reduce verification effort.
Reasoning about correctness and performance; optional: distributed systems, model checking
Real-time localisation and mapping in GPS-denied environments under a 5 W power cap, using spiking neural networks on neuromorphic hardware.
Computer architecture basics, HDL, some ML

Research, product design and scaling of assistive technologies - navigation, braille displays, tactile graphics and AI-based assistants - for persons with disabilities.

An open technology stack that turns satellite, drone and hydrology data into actionable insights and tools that help rural communities plan for water security and sustainable landscapes.
System building, data science, ML, AI application engineering
Network of development and ecological-restoration organisations
Extract information from research papers to create domain-specific knowledge graphs, run ecological data through the graph to reason about specific stresses, and build plans.
Refer to the owncloud document.
Refer to the owncloud document.
Coupling AI models with dataflow analysis and formal methods to find safety and security defects, optimise performance and transpile to quantum circuits.

A new kind of 3D design tool that combines AR, VR, haptics and LLMs so engineers can 'design 3D in 3D'.

Fabricating computing devices from protein polymers such as microtubules - reading, storing and converting photochemical information for green computing.

Making regulatory compliance a first-class citizen in how data is stored, processed and queried.
Cancer detection, surgical video intelligence and driver assistance, deployed with AIIMS Delhi, PGI Chandigarh, Honda, Yamaha and others.
AIIMS Delhi, PGI Chandigarh, Honda, Yamaha
Abstractions that let programmers trade off runtime, cost and accuracy in LLM workflows, and make them incremental.
Python, systems performance, SQL; optional: dataflow systems, provenance, distributed systems
R Systems CoE on Sustainable AI

Next-generation memory systems, AI/ML hardware and 3D chip stacking, developed in close collaboration with industry for integration into products.

Privacy-preserving search and intelligent indexing so valuable datasets can be found across organisations without centralising them, in line with GDPR and India's DPDP Act.

Geospatial machine learning and statistical computing to locate pockets of under-vaccination and explain their behavioural and social drivers.
University of Montreal, London School of Hygiene and Tropical Medicine
Open, reproducible, medical-grade models on ECG and PPG data - atrial fibrillation, rhythm monitoring, non-invasive glucose estimation, stroke-type detection.
ML/DL, programming, interest in medical applications
AIIMS Delhi
Mapping classical Boolean functions through reversible logic synthesis to efficient circuits for 2-D nearest-neighbour qubit arrays.
C/C++, compilers, back-end optimisation, physical CAD, LLVM
A co-designed streaming pipeline that calls species from raw nanopore signal in under 60 seconds at under 30 W, for point-of-care diagnostics.

ML-based intrusion detection for cyber-physical systems, lightweight authentication for constrained devices, and privacy-preserving techniques for sensitive data.
A container abstraction that lets a verifier independently reproduce experiments that depend on remote services such as LLM APIs.
Operating systems; optional: seccomp, confidential computing