Research programs
IAIRO wiki
A curated view of ongoing and completed work across enterprise and foundation AI.
IAIRO is pioneering new types of sovereign AI for India. From Consumer AI to Enterprise AI. From Big AI to Small AI.
7
Active projects
0
Archived resources
Pharma, Health, Sustainability, and Education
Research focus
Specialized systems, grounded AI, and applied research
Reusable project cards keep the site modular while the visual language stays consistent across the research portfolio.
SAMVIT Platform
India's sovereign AI platform for specialized foundational models that are frugal, secure, and deployable across cloud, enterprise, edge, and future mobile environments.
Active Projects
NIDHI: Explainable Suicide Risk Detection
1st-place system (out of 55 teams) in the IEEE BigData 2026 Cup's Explainable Suicide Risk Detection challenge — combining task-specific modeling with knowledge-guided approaches to assess suicide risk from Reddit posts with grounded evidence and interpretable explanations.
Smart Pilot for Drug Manufacturing
The Pharma Manufacturing Intelligence Platform is an AI-driven system designed to improve monitoring, diagnosis, forecasting, and operational decision-making in pharmaceutical manufacturing environments. The platform combines anomaly detection, causal root cause analysis, predictive forecasting, and document-based knowledge retrieval into a single integrated workflow.
Context Preservation in Long-Form Video Generation via Scene Graphs
Scene-graph-conditioned video continuation to reduce context drift, hallucination, and identity loss in long-form generation.
Project MEDHA — Mathematical Engine for Domain-specific Higher-order Autoformalisation
A neuro-symbolic Small Language Model (SLM) for translating natural language into Lean 4 formal specifications, enabling verification-native autonomy for mission-critical domains such as finance, semiconductors, and robotics.
Drug Repurposing Platform
A neurosymbolic drug-repurposing framework that integrates heterogeneous biomedical knowledge graphs into a unified knowledge representation. The framework combines symbolic KG retrieval through SPARQL/Cypher with sAIM-based GNN link prediction to identify both established and potentially novel drug–disease associations. Candidate associations are ranked by combining KG-derived evidence with predictive confidence, while each prediction is accompanied by supporting reasoning paths, provenance, confidence scores, and human-readable explanations to enable transparent and interpretable drug-repurposing decisions.
Long-Range Embodied QA: Benchmarking and Memory-Augmented Navigation in Multi-Room VLM Agents
A procedurally generated benchmark for embodied question answering with controllable room-count horizon, paired with a KG-augmented episodic memory system showing that structured trajectory memory improves answering accuracy by up to 6.6 points on route-history questions.
PAL: Personal Adaptive Learner
AI-powered platform transforming lecture videos into interactive learning experiences with real-time adaptive questioning.
VLMs, Alignment, Safety & Multimodal Knowledge Graphs
VLM hallucination benchmarking, multimodal safety, AI-generated content detection, and knowledge-graph-grounded vision-language reasoning. Foundation work for recent developments in vision-language model safety and grounding.
Archived Projects and Resources
No archived projects found.
Publications
All IAIRO publications browse grouped by research project, or in one reverse-chronological timeline.
Long-Range Embodied QA: Benchmarking and Memory-Augmented Navigation in Multi-Room VLM Agents
Embodied AI / Vision-Language Models / Benchmarking
.
2026 - Present
2026
Answering Without Arriving: An Evidence Audit for Embodied Question Answering
Shubh Garg, Divyansh Pandey, Niyati Rawal, Vedant Khandelwal, Amit Sheth
Project MEDHA — Mathematical Engine for Domain-specific Higher-order Autoformalisation
Formal Methods & Verifiable AI
.
2026 - Present
2026
SEMINAR: Synchronization Engine for Mathematical Illustration, Narration, and Reasoning
Rishov Mondal, Amit Sheth
Smart Pilot for Drug Manufacturing
Smart Manufacturing, Pharma
.
2026 - Present
2026
CausalPulse: An Industrial-Grade Neurosymbolic Multi-Agent Copilot for Causal Diagnostics in Smart Manufacturing
Shyalika, C., Jaimini, U., Henson, C. and Sheth, A.
IndustryAssetEQA: A Neurosymbolic Operational Intelligence System for Embodied Question Answering in Industrial Asset Maintenance
Shyalika, C., Patel, D.C. and Sheth, A.
NIDHI: Explainable Suicide Risk Detection
Explainable AI for Mental Health
.
2026
2026
Bag of Tricks or Bag of Myths? Reducing Modeling Complexity with Task Knowledge in Explainable Suicide Risk Assessment
Shlok Shelat, Shrey Salvi, Souvik Roy, Manas Gaur, Amit Sheth
Sustainability
2026
AI-enabled water resilience: Applications and future directions
Niyogi, D., Tiwari, A., Patel, A. and Neupane, B.
Others
2026
Abstention vs. Hallucination: Benchmarking LLM Source Attribution for Scientific Citations
Deepa Tilwani, Yash Saxena, Sevedali Mohammadli, Ankur Padia, Edward Raff, Amit Sheth, Srinivasan Parthasarathy, Manas Gaur
SOFAI-DT: Return-Conditioned Control for Efficient Verified Reasoning
Rahul Maity, Vedant Khandelwal, Amit Sheth
MED-EVAL: Clinical Fact Alignment for Efficient Medical Reasoning
Rishit Agarwal, Vedant Khandelwal, Amit Sheth
Neuro-symbolic PRM: Enhancing Scientific Reasoning via Structured Traces and Symbolic Verification
Zi, Y., Xu, C., Bhattacharya, S., Foltin, M., Sheth, A.
Cognitive Fatigue in Autoregressive Transformers: Formalization and Measurement
R. Marwah, R. Garimella, V. Pallagani, A. Jain, M. Stewart, A. Sheth
One Size Does Not Fit All: Revisiting World Models and Neurosymbolic AI
Amit Sheth, Madhur Thareja, Anushka Pawar, Niyati Rawal
LiveSpec: A Tiered Runtime for Planning under Live-Editable Constraints
Vedant Khandelwal, Prateek Biswas, Amit Sheth
Where Should Knowledge Enter? A Layered Framework for Knowledge Infusion in Multimodal Iterative Generative Model
Renjith Prasad, Chathurangi Shyalika, Anushka Pawar, Amit Sheth
Toward Neurosymbolic Reinforcement Learning via Editable Specifications
Vedant Khandelwal, Hong Yung Yip, Amit Sheth
CAMEO: A Black-Box Behavioral Diagnostic for Evaluation-Aware Concealment in Language Models
Vatsalya Soni, Shaivi Malik, Vedant Maheshwari, Venkatesan Nadimuthu, Yash Kulkarni, Amit Sheth
2026
CAMEO: A Black-Box Behavioral Diagnostic for Evaluation-Aware Concealment in Language Models
Vatsalya Soni, Shaivi Malik, Vedant Maheshwari, Venkatesan Nadimuthu, Yash Kulkarni, Amit Sheth
Toward Neurosymbolic Reinforcement Learning via Editable Specifications
Vedant Khandelwal, Hong Yung Yip, Amit Sheth
Where Should Knowledge Enter? A Layered Framework for Knowledge Infusion in Multimodal Iterative Generative Model
Renjith Prasad, Chathurangi Shyalika, Anushka Pawar, Amit Sheth
LiveSpec: A Tiered Runtime for Planning under Live-Editable Constraints
Vedant Khandelwal, Prateek Biswas, Amit Sheth
One Size Does Not Fit All: Revisiting World Models and Neurosymbolic AI
Amit Sheth, Madhur Thareja, Anushka Pawar, Niyati Rawal
Cognitive Fatigue in Autoregressive Transformers: Formalization and Measurement
R. Marwah, R. Garimella, V. Pallagani, A. Jain, M. Stewart, A. Sheth
Neuro-symbolic PRM: Enhancing Scientific Reasoning via Structured Traces and Symbolic Verification
Zi, Y., Xu, C., Bhattacharya, S., Foltin, M., Sheth, A.
MED-EVAL: Clinical Fact Alignment for Efficient Medical Reasoning
Rishit Agarwal, Vedant Khandelwal, Amit Sheth
SOFAI-DT: Return-Conditioned Control for Efficient Verified Reasoning
Rahul Maity, Vedant Khandelwal, Amit Sheth
Abstention vs. Hallucination: Benchmarking LLM Source Attribution for Scientific Citations
Deepa Tilwani, Yash Saxena, Sevedali Mohammadli, Ankur Padia, Edward Raff, Amit Sheth, Srinivasan Parthasarathy, Manas Gaur
AI-enabled water resilience: Applications and future directions
Niyogi, D., Tiwari, A., Patel, A. and Neupane, B.
Bag of Tricks or Bag of Myths? Reducing Modeling Complexity with Task Knowledge in Explainable Suicide Risk Assessment
Shlok Shelat, Shrey Salvi, Souvik Roy, Manas Gaur, Amit Sheth
Answering Without Arriving: An Evidence Audit for Embodied Question Answering
Shubh Garg, Divyansh Pandey, Niyati Rawal, Vedant Khandelwal, Amit Sheth
SEMINAR: Synchronization Engine for Mathematical Illustration, Narration, and Reasoning
Rishov Mondal, Amit Sheth
IndustryAssetEQA: A Neurosymbolic Operational Intelligence System for Embodied Question Answering in Industrial Asset Maintenance
Shyalika, C., Patel, D.C. and Sheth, A.
CausalPulse: An Industrial-Grade Neurosymbolic Multi-Agent Copilot for Causal Diagnostics in Smart Manufacturing
Shyalika, C., Jaimini, U., Henson, C. and Sheth, A.
Let's Collaborate
Join the network behind India's AI future
Partner with IAIRO or subscribe to stay updated on research, events, and opportunities.