Python TypeScript JavaScript Java PyTorch Hugging Face LangChain LangGraph LlamaIndex Next.js React Tailwind CSS FastAPI Node.js RAG LLM Optimization PEFT Agent Systems Evals PostgreSQL MongoDB Qdrant Docker Vercel

3 publications/preprints  ·  15x hackathon wins  ·  ML Researcher @ Raapid  ·  SWE Intern @ C3alabs

Currently ML Researcher @ Raapid Inc + SWE Intern @ C3alabs — building across LLM optimization, RAG, agents, evals, and production AI systems.

• AgentForge — Published Python terminal AI coding-agent harness with typed tools, approval gates, MCP tools, subagents, checkpoints, event logs, restore flows, and a Rich TUI
• MemexLLM — Deployed RAG platform with hybrid semantic + BM25 retrieval, Cohere reranking, Qdrant vector search, LlamaIndex orchestration, citations, and confidence thresholds
• BlinkAI — Privacy-first desktop AI assistant with voice input, screen context, memory, MCP/Composio tool routing, and workflow automation
• Artificial Guruji — AI-powered exam-prep platform for personalized study plans, flashcards, quizzes, and learning workflows
• Paper-Replicating — Converting ML papers and research ideas into working notebooks and Python implementations
• DSA — Data structures and algorithms in Java

• GRIT: Geometry-Aware PEFT with K-FAC Preconditioning, Fisher-Guided Reprojection, and Dynamic Rank Adaptation

  • Co-authored · arXiv preprint
  • Focused on parameter-efficient LLM adaptation, benchmark evaluation, ablations, and comparison against full fine-tuning and LoRA
  • Google Scholar ↗

• A Comprehensive Review of Microsoft Semantic Kernel

  • IEEE DELCON 2025 conference publication
  • Covers Semantic Kernel architecture, capabilities, agent-oriented patterns, orchestration, and enterprise AI research directions

• AoT-BTS: Explainable Age-of-Trust Based Behavioral Scoring

  • Research Square preprint
  • Explainable real-time zero-trust scoring with SDN, behavioral signals, and SHAP-based interpretation

I build from zero. Whether it's implementing transformers from the paper, wiring agent systems into production workflows, or co-developing research around PEFT optimization — I work from first principles to shipped systems. Research tells me what's possible. Building tells me what's true.

• LLM optimization — PEFT, K-FAC preconditioning, dynamic rank adaptation, benchmark-driven evaluation
• RAG systems — hybrid retrieval, reranking, citations, confidence thresholds, long-context document reasoning
• AI agent systems — tool orchestration, MCP, approval gates, persistence, evals, and multi-agent patterns
• Production AI — structured outputs, memory, audit trails, human-in-the-loop review, regression tests, fallback paths

I'm open to research collaborations, internships, and full-time roles in ML engineering, AI systems, agent infrastructure, RAG, or full-stack AI development.

If you're building production AI — agents, RAG, evals, or LLM infrastructure — I work from paper to deployment and ship systems that can be inspected, tested, and improved.