Research & Publications
Work at the intersection of peer-review transparency, argumentation theory, and applied GenAI systems.
Dynamic Optimization of Peer Review Length Using Information Density Analysis
Developed a heuristic framework based on Cognitive Load and Discourse Analysis to optimize peer review lengths using information density, content relevance, argument strength, and readability metrics.
AntiBug: Runtime-Aware Multimodal Agentic Program Repair for Production Applications
Built a 12-agent LangGraph pipeline for runtime-aware program repair, integrating a Next.js telemetry SDK, stack traces, user-event context, and DOM screenshots for root-cause diagnosis and automated patch synthesis. Evaluated on 69 real-world bugs, fixing 32/69 issues vs 9/69 for GPT-4o.
FRAC-MAS: A Safe and Explainable Multi-Agent System for Fracture Diagnosis
Engineered a multi-agent clinical decision support system that integrates a 4-model vision ensemble (MaxViT, RAD-DINO, YOLO, DenseNet) with specialized LLM agents and delivers verifiable, RAG-grounded reasoning. Implemented conformal prediction for statistically grounded differential diagnoses, achieving 92.0% empirical coverage.
DigniFy: Multi-Modal Multilingual Online Hate Speech Detection
Presents a theoretical framework and pipeline for multilingual, multimodal safety systems with interactive LLM orchestration to detect online hate speech.
Evaluating Large Language Models for Automated Requirement Generation in IT Projects
A comprehensive benchmark comparing DeepSeek R1, GPT-4o, Gemini 2.0, and LLaMA 3.2 on BLEU, Levenshtein, and Jaccard metrics for requirement engineering.
Achievements & Recognition
GATE DS&AI Rank 499
Ranked 499 nationwide among 60,000+ candidates in GATE Data Science & AI, demonstrating strong theoretical foundations.
Amazon ML Challenge Rank 292
Placed 292 out of 70,000+ participants in Amazon ML Challenge 2024, showcasing applied machine-learning expertise.
CodeBounty 2024 Winner
Won the flagship competitive programming contest at DJSCE, highlighting algorithmic design strengths.