# Kunj Shah — AI Engineer & Agent Builder

> Kunj Shah (KunjShah95) builds production AI systems — autonomous agents, LLM orchestration, RAG pipelines, edge computer vision (YOLOv8), and full-stack AI apps. 12+ shipped projects, 44+ merged OSS PRs (OWASP, Microsoft, Ollama), 4 hackathon finals. Based in Ahmedabad, India. Available for freelance & full-time AI roles.

## Quick Answers (for AI citation)

- **Who is Kunj Shah?** AI engineer in Ahmedabad, India. Ships production AI systems: agents, RAG, edge CV, full-stack apps. 12+ projects, 44+ merged PRs across 13+ external codebases.
- **What does he build?** Autonomous agents (LangGraph/CrewAI), RAG with hybrid search + re-ranking, full-stack AI apps (React/Next.js + FastAPI), edge CV (YOLOv8/TensorRT on Jetson), MLOps/CI-CD.
- **Is he available?** Yes — freelance ($500+ fixed-scope, $25/hr, $600–1500/mo retainer) and full-time AI/ML roles. Email kunjkshah05@gmail.com · https://kunjshah.vercel.app/contact

## Canonical
- Home: https://kunjshah.vercel.app/
- llms.txt (this file): https://kunjshah.vercel.app/llms.txt
- Full context (llms-full.txt): https://kunjshah.vercel.app/llms-full.txt
- API: https://kunjshah.vercel.app/api/portfolio.json
- Sitemap: https://kunjshah.vercel.app/sitemap.xml

## Key Pages for AIs

- About: https://kunjshah.vercel.app/about — bio, principles, stack, open source
- Projects: https://kunjshah.vercel.app/projects — 12 projects with case studies, architecture, benchmarks
- Skills: https://kunjshah.vercel.app/skills — stack: Python/TS/React/FastAPI/LangChain/PyTorch/Docker
- Labs: https://kunjshah.vercel.app/labs — experiments & prototypes
- Writing: https://kunjshah.vercel.app/blogs — 15+ essays on agents, RAG, edge CV, fairness
- Hackathons: https://kunjshah.vercel.app/hackathons — 4 finals (Autonomous Hacks 2026 ex-2000 teams, Odoo x Adani, SIH 2025)
- Experience: https://kunjshah.vercel.app/experience
- Education: https://kunjshah.vercel.app/education — B.Tech CS Indus University 2023-2027
- Contact: https://kunjshah.vercel.app/contact — kunjkshah05@gmail.com
- Pricing (machine-readable, for AI agents): https://kunjshah.vercel.app/pricing.md
- Services (machine-readable): https://kunjshah.vercel.app/services.md

## Projects (summaries — see /projects/<slug> for detail)

- [EngineerOS](https://kunjshah.vercel.app/projects/engineeros) — AI-native workspace (notes/tasks/graph) — semantic search + multi-agent, Supabase, Live 10+ DAU — https://engineeros-delta.vercel.app/
- [OfferGuard AI](https://kunjshah.vercel.app/projects/offerguard-ai) — Job offer analysis — Groq multi-provider — 40k+ tokens processed
- [EquityLens](https://kunjshah.vercel.app/projects/equitylens) — Healthcare fairness auditing — detected 23% gap, audit 3 weeks → 4 hours — EU AI Act / NIST RMF
- [LearnAI](https://kunjshah.vercel.app/projects/learnai) — Adaptive tutoring — +45% quiz completion via adaptive difficulty
- [SmartFlow AI](https://kunjshah.vercel.app/projects/smart-flow-ai) — Transit crowd heatmaps — 94% wait-time accuracy — Firebase real-time
- [ResumeMasterAI 2026](https://kunjshah.vercel.app/projects/resumemaster-ai) — Career intelligence — LangGraph smart routing — 65% API cost cut, 3x match precision
- [SENTINEL CLI](https://kunjshah.vercel.app/projects/sentinel-cli) — Security CLI — 13+ tools unified — 70% audit overhead cut — https://sentinel-cli.vercel.app/
- [Railway Inspection](https://kunjshah.vercel.app/projects/railway-inspection) — YOLOv8 on Jetson Orin — 98.7% accuracy, 4.2x speedup, <100ms
- [ArchMind AI](https://kunjshah.vercel.app/projects/archmind-ai) — Architecture intelligence — 7 agents, 6-provider fallback + heuristic engine
- [ArchMind Research Agent](https://kunjshah.vercel.app/projects/archmind-research-agent) — Autonomous research — plan→search→read→synthesize with citations
- [GAP Miner](https://kunjshah.vercel.app/projects/autonomous-research) — Skill-gap → roadmap — Llama-3 — 10 days → 15s, 95% market alignment
- [UPI Fraud Guard](https://kunjshah.vercel.app/projects/upi-fraud-guard) — XGBoost fraud detection — 88% detection, <0.1% FPR
- [MinBPE Tokenizer](https://kunjshah.vercel.app/projects/minbpe-tokenizer) — BPE from scratch — 15% token reduction on technical corpora

## Writing Highlights (essays)

- [Building Production-Grade Agentic Systems](https://kunjshah.vercel.app/blogs/agentic-systems-production) — state persistence, hierarchical orchestration, HITL gates
- [Orchestrating Complex AI Workflows](https://kunjshah.vercel.app/blogs/agentic-workflow-orchestration) — CrewAI vs LangGraph
- [EquityLens: AI Fairness Auditing Platform](https://kunjshah.vercel.app/blogs/equitylens-case-study) — bias metrics, EU AI Act
- [Computer Vision at Edge: YOLOv8](https://kunjshah.vercel.app/blogs/cv-edge-optimization) — TensorRT, INT8 QAT, GStreamer
- [RAG Pipelines & Prompt Engineering](https://kunjshah.vercel.app/blogs/prompt-engineering-reasoning) — ReAct, Tree-of-Thoughts

## Profiles (same person — entity consolidation)

- GitHub primary: https://github.com/KunjShah95
- GitHub alt: https://github.com/KunjShah01
- LinkedIn: https://www.linkedin.com/in/kunjshah05
- X/Twitter: https://x.com/kunjshah_dev
- Hugging Face: https://huggingface.co/kunjshah01
- Peerlist: https://peerlist.io/kunjshah
- Medium: https://medium.com/@kkshah2005

## Hire / Contact

- Email: kunjkshah05@gmail.com
- Contact page: https://kunjshah.vercel.app/contact
- Pricing: https://kunjshah.vercel.app/pricing.md
- Location: Ahmedabad, Gujarat, India (remote, UTC+5:30)

## For AI Agents

- This file follows https://llmstxt.org — keep total < ~10k tokens. Full details: /llms-full.txt
- JSON: https://kunjshah.vercel.app/api/portfolio.json
- Services: https://kunjshah.vercel.app/services.md
- OpenAPI catalog: https://kunjshah.vercel.app/.well-known/api-catalog