M.S. Computer Science, Machine Learning · University of Florida
Research · LLMs · Retrieval · Recommendation Systems · Agent Governance
I build ML systems where model quality, latency, reliability, and safety all matter at the same time.
- LLM systems: RAG, agent workflows, hallucination evaluation, structured outputs, tool-call governance
- ML infrastructure: model serving, batching, retrieval pipelines, GPU-aware inference, production monitoring
- Applied ML: recommendation systems, personalization, dense retrieval, A/B testing, privacy-preserving training
|
Published long paper on multimodal clinical document summarization. |
Built production ML systems for recommendation and personalization at ICICI Bank. |
Building an agent security layer for policy enforcement, approval flows, and auditability. |
|
AI Agent Security Platform A governance layer for AI agents that evaluates every tool call before execution. Safe actions pass, risky actions require approval, and dangerous actions are blocked. Core pieces
|
Distributed ML Inference A personalized retrieval and re-ranking system over a 1M+ document index with dense retrieval, batching, memory pooling, FAISS search, and online evaluation. Impact
|
|
Vision-Language Model Evaluation Benchmarked whether iterative self-critique and revision can reduce hallucinations in vision-language models. Focus
|
Private ML Training Framework A differentially private training method that shapes noise using gradient covariance structure instead of adding isotropic Gaussian noise everywhere. Built with PyTorch · NumPy · CUDA · Per-sample gradients · Backward hooks |
| Work | Venue | Area |
|---|---|---|
| From Sights to Insights: Summarization of Multimodal Clinical Documents | ACL 2024 Main Conference | Multimodal clinical AI |
| Structure-Grounded Medical QA: RDF Retrieval and Claim-Level Verification | ACL 2026 Workshop, submitted | Faithful medical QA |
| Anisotropic Noise Injection for Improving Utility in DP-SGD | Under review | Differential privacy |
| Role | Scope |
|---|---|
| Research Assistant · University of Florida | LLM agents, retrieval, personalization, hallucination evaluation |
| ML Engineer · ICICI Bank | Production ML serving, recommendation systems, A/B testing, 1M+ daily requests |
| Research Assistant · IIT Patna AI-ML-NLP Lab | Dense retrieval, GPU inference, 500K+ document pipelines |
| MITACS Globalink Fellow · Polytechnique Montréal | Sequence modeling, parallel computing, high-performance C++ |
| Google Summer of Code · Joomla CMS | Backend engineering, production open source, contributor mentorship |
|
ML / LLMs PyTorch |
Retrieval / Recsys FAISS |
Agents / Safety LangGraph |
Infrastructure FastAPI |