Research AI Projects

Explore advanced AI research ideas, publications, experimental methodologies and innovative solutions.

Finance AI
13. Graph-Based Fraud Ring Detection in Synthetic Banking Networks

Can graph features detect coordinated fraud better than transaction-only models?

Finance AI
14. Stress Testing AI Models for Indian SME Credit Risk

How do credit-risk models behave under economic stress scenarios and missing informal-economy signals?

Finance AI
15. Robust Backtesting Framework for Student Trading Strategies

How often do student trading strategies fail after transaction costs, slippage, and regime changes?

Hardware AI
16. AI Surrogate Models for Hardware Thermal Simulation

Can a neural surrogate approximate thermal simulation fast enough for design exploration?

Hardware AI
17. Fault Prediction from Simulated Digital Signal Traces

Can AI classify early hardware faults from noisy signal patterns before failure thresholds are crossed?

Weather AI
18. Extreme Rainfall Early Warning Using Hybrid ML and Simulation

Can data-driven models improve short-term rainfall risk alerts when historical data is sparse?

Weather AI
19. Urban Heat Island Modeling with Remote Sensing and Weather Signals

Which satellite and weather features best explain urban heat patterns in Indian cities?

Healthcare AI
20. Uncertainty-Aware Medical Image Screening for Student Research Labs

How can screening models communicate uncertainty so non-clinical users do not overtrust outputs?

Healthcare AI
21. Clinical RAG Safety Evaluation with Synthetic Patient Cases

What retrieval and refusal strategies reduce unsafe healthcare answers?

Agriculture AI
22. Multimodal Crop Advisory Combining Images, Weather, and Text Manuals

Can multimodal evidence improve farm recommendations compared with image-only plant disease models?

Agriculture AI
23. Low-Cost Drone Crop Health Mapping for Small Farms

Can low-resolution drone imagery provide actionable crop stress maps for small farms?

Cybersecurity AI
24. LLM-Assisted SOC Triage with Evidence-Ranked Alerts

Can an LLM summarize alerts while preserving evidence and reducing analyst workload?