AI Researcher & Engineer working on one question: when an AI model explains its answer or describes a medical image, can we trust what it says?
My research focuses on explanation faithfulness and hallucination detection in large language and vision-language models, with an emphasis on medical imaging.
- AI Engineer at ZenteiQ.ai, building scientific foundation models
- M.Tech at IIIT Dharwad, advised by Dr. Shirshendu Layek
-
What Does Answer Change Rate Actually Measure? A Specificity Audit of Causal Intervention Tests for Explanation Faithfulness Durgesh Rao, Shirshendu Layek · GroundLM Workshop @ EMNLP 2026 [Paper] [Code]
-
Efficient Deep Learning Model for Analyzing Muscle Activity Patterns in Biomechanical Simulations Dharmendra Dangi, Dheeraj Kumar Dixit, Amit Bhagat, Durgesh Rao, Jeetendra Kumar Gupta · SN Computer Science (Springer), 2025 [Paper]
| Project | Description |
|---|---|
| Hard Reasoning Benchmark | Reasoning benchmark built by filtering problems with model-disagreement scores |
| Llama-8B GRPO | RL fine-tuning of Llama-8B-Instruct on Python code with Group Relative Policy Optimization |
| MedGemma Medical VLM | LoRA fine-tuning of MedGemma 1.5 on CT, MRI and X-ray images, with a Gradio demo |
| Retinal Diagnostic VLM | Qwen2.5-VL-7B fine-tuned with LoRA and medical distillation for retinal diagnosis |
| Fake Instagram Profile Detection | Neural network classifier for fake social media profiles |
- DRDO (DEBEL), Bengaluru · Graduate Research Fellow, biomedical engineering
- DIAT, Pune · Anomaly detection on maritime traffic data with the Indian Navy
- ISRO Space Applications Centre · Collocating INSAT-3D satellite data with NASA's CERES
- MANIT Bhopal · Deep learning for sports biomechanics
- Kaggle Discussions Grandmaster · Notebooks Master · Datasets Expert
- Winner, Kavach Cybersecurity Hackathon (BPR&D, Ministry of Home Affairs)
Python · PyTorch · Hugging Face Transformers · PEFT / LoRA · TensorFlow · scikit-learn · OpenCV · GCP · Azure