Agentic AI & MCP Multi-agent orchestration, custom MCP servers/tools, scoped tool access, agent evaluation harnesses

RAG / GraphRAG Hybrid graph + vector retrieval, LlamaIndex, Neo4j, FAISS, query routing, retrieval-accuracy tuning

LLM Systems Anthropic Claude API, OpenAI, Azure AI Foundry, prompt engineering, cost/latency optimization

Backend & APIs FastAPI, Django REST Framework, Node.js, clean architecture, idempotent payment systems

Data & Streaming Kafka/Redpanda event pipelines, ETL, PostgreSQL, Redis, MongoDB

MLOps & Cloud Docker, Kubernetes (Rancher/CaaS), GitHub Actions CI/CD, AWS (Lambda, EC2, SAM), Azure, GCP

Artificial Intelligence Engineer · CobuildX.ai · _{Jun 2026 – Present}

• Shipped AI custom agents with MCP tool integration, giving each agent scoped GitHub repo access for code-aware automated reviews
• Built full-stack features (React + TypeScript, FastAPI on AWS Lambda) including a real-time collaborative Kanban board
• Audited production Supabase/Postgres row-level security across a multi-tenant data model
• Built a config-driven entity-extraction evaluation framework, surfacing 3 production defects with zero code changes

Graduate Technical Intern — AI Engineering · Intel Technology India · _{Jul 2025 – Feb 2026}

• Designed a production multi-agent LLM ticket-automation system (Azure AI Foundry + ServiceNow) — 35–45% lower model cost, 39.4% fewer input tokens
• Built a Retriever API over internal wikis — 12s avg response, 1.8×–2.3× throughput gain
• Engineered a GraphRAG platform (LlamaIndex, Neo4j, FAISS hybrid search) — +32% retrieval accuracy, -45% lookup latency
• Ran an MLOps pipeline (Docker, Kubernetes/Rancher, GitHub Actions) with zero-regression releases

Adaptive Multi-Agent Generative AI System for Automated Legal Contract Drafting and Advisory _{(Filed, 2025)} Multi-agent architecture (Clause Selector, Drafter, Compliance, Verifier) with cryptographic provenance logging — 92% F1 on document understanding, 42% faster NDA drafting

Hardware-Accelerated SAR Semantic Downlink Onboard VLM-based semantic processing for satellite imagery — cuts bandwidth usage 95%+ with veracity enforcement

Meta FAISS — PR #4792 — Fixed a multi-platform CI regression, restoring build stability for a vector-search library used at global scale Google DeepMind Gemma — PR #590 — Resolved a runtime crash and added regression tests for a production LLM inference pipeline

Agent Tool Gateway (FastAPI + Pydantic) — <100ms dispatch latency, a "Safe SQL" validator blocking 100% of injection risk, and an automated tool-selection eval harness. _{Python FastAPI MCP Pydantic}

Event-driven IoT pipeline on Kafka/Redpanda processing 5,000+ events/sec across 45+ simulated devices with exactly-once processing and real-time anomaly detection. _{Python Kafka FastAPI}

Production-grade RAG system for enterprise compliance and audit-safe policy reasoning. _{Python RAG LLM}

Production ML system for defect detection in steel hot rolling — FastAPI + 6-model ensemble + Jenkins CI/CD + Prometheus + Grafana. _{Python FastAPI MLOps}

Payments API with a Redis-backed idempotency engine guaranteeing zero duplicate charges under concurrent retries and network failures. _{Python FastAPI Redis}

XGBoost credit-default model (81.7% accuracy, 0.77 ROC-AUC) deployed dual-strategy on AWS Lambda + Dockerized EC2 with sub-150ms latency. _{Python XGBoost AWS}

Niral Thiruvizha 2024 (Govt. of Tamil Nadu) — AgroVoiceAI ranked Top 100 statewide (from 500+ shortlisted) — AI-driven voice-based agricultural assistance Winner — Praestantia 2023 — StockistAI (Hybrid LSTM + SVM stock trend forecasting) Winner — Technoteen 2021 (LICET) — DigiPass (Flutter + QR-based digital bus pass system) Ashok Leyland Hackathon Finalist — IntelliRoute (AI-powered EV energy management system) Organizer — Hackwarts Hackathon (Xplore 2023, LICET) Sports Captain (CSE Dept., LICET) & CM Trophy Winners (VIT 2024, District Level)