Nisarg P.

MOST PROJECTS DON'T FAIL AT THE IDEA STAGE — THEY FAIL AT PRODUCTION Top Rated | 100% Job Success | 12+ Years in Enterprise Software 𝗖𝗿𝗲𝗮𝘁𝗼𝗿 𝗼𝗳 𝗖𝗼𝗮𝗹𝗲𝗻𝘁 — 𝗼𝗽𝗲𝗻-𝘀𝗼𝘂𝗿𝗰𝗲 𝗰𝗼𝗴𝗻𝗶𝘁𝗶𝘃𝗲 𝗰𝗮𝗰𝗵𝗲 𝗳𝗼𝗿 𝗔𝗜 𝗮𝗴𝗲𝗻𝘁𝘀 & 𝗥𝗔𝗚 AI or full-stack — same story: the demo works, then real usage breaks it. I've taken 6+ enterprise AI systems from prototype → live deployment across pharma, energy, government, and financial sectors — and I've built the full-stack applications around them (Java/Spring Boot, React/Next.js, enterprise portals). If you have a POC hitting real-world constraints (latency, cost, scaling, reliability) — or need a full application built to hold in production — this is exactly where I work. WHAT I ACTUALLY BUILD 1️⃣ Production-grade agentic AI — not chatbot wrappers • Multi-agent orchestration • Enterprise knowledge pipelines (RAG) • Real-world constraints: latency, cost, security, scale 2️⃣ Full-stack enterprise applications • Spring Boot / FastAPI backends, REST APIs, microservices • React / Next.js / TypeScript frontends • Enterprise portals, dashboards, data platforms This is not about connecting an LLM to your data — or shipping another CRUD app. This is about building systems that hold under real usage. TECH STACK (2025–26) 🔹 Agent Frameworks MCP (Model Context Protocol), Google A2A, LangGraph, CrewAI, Azure AI Agent Service 🔹 RAG (beyond basics) Hybrid search, query rewriting, role-based access, eval loops, retrieval benchmarking 🔹 LLM Orchestration GPT-4o, Gemini, Claude — tool use, function calling, streaming, cost control 🔹 Backend Engineering Java, Spring Boot, Python, FastAPI, Flask — high-performance AI APIs, async pipelines, long-running job orchestration 🔹 Frontend Engineering React, Next.js, TypeScript — dashboards, admin panels, SSR + static sites 🔹 Voice AI & Real-Time Systems Speech-to-Text (STT), Text-to-Speech (TTS), real-time voice agents, streaming pipelines (Agora, WebRTC) 🔹 Enterprise Platforms Liferay DXP — portals, headless APIs, custom components 🔹 Cloud & Data GCP (Cloud Run, BigQuery, Pub/Sub), Azure AI Search, Document Intelligence, PostgreSQL, CI/CD, Docker REAL PRODUCTION PROBLEMS I'VE SOLVED Not "skills" — problems I've already been hired to fix: • 4+ hour AI jobs surviving GCP evictions (checkpointing + state recovery) • Multi-agent RAG with strict role-based document access • Reduced latency from 12s → under 3s (P99) via pipeline redesign • Passed enterprise Azure security compliance (pharma platform) • Reduced hallucinations via eval loops + confidence gating • Fixed token cost overruns via batching + model routing • High-volume API ingestion — rate limits, concurrency control, Pub/Sub + BigQuery • Multi-site enterprise portals with role-based content (global pharma) SELECTED PROJECTS 1. Marketplace Data & AI Platform (Ongoing) AI + full-stack, live: Spring Boot + GCP ingestion of Amazon seller APIs → BigQuery. Pub/Sub, rate limiting, concurrency control, AI insights layer. 2. Government AI Recruiter (US) 7-agent pipeline for candidate scoring + reporting. Reduced screening effort by 60–70%. 3. Coalent — Open-Source Cognitive Cache (on GitHub & PyPI) Caches the understanding your LLM builds — not raw chunks — reuses it across queries and agents, and stays fresh when sources change. RAG-level accuracy at ~1/3 the context tokens. Apache-2.0. pip install coalent 4. Enterprise Intranet Platform (Japan) Full-stack portal for a global pharma client — custom components, headless APIs, multi-site workflows. 5. Pharma Oncology AI Platform (Japan) Hybrid RAG + role-based access + streaming. Passed enterprise security review and reduced internal support load. REAL-TIME AI & VOICE SYSTEMS I build conversational AI systems beyond text — including real-time voice interaction and AI-driven interviews. • AI interview platforms with live questioning, evaluation, and scoring • Voice agents using STT + TTS with sub-3s response pipelines • Streaming-based architectures (WebRTC / Agora) for real-time interaction • Integration with LLM reasoning pipelines for dynamic conversations These demand tight control over latency, orchestration, and response quality — not just model integration. WHO I WORK WITH ✔ Teams with a real use case (not "we want AI") ✔ Projects beyond POC or serious about production ✔ Teams that need both AI + infrastructure expertise ✔ One engineer for the AI and the app around it ✔ Clients who care about accuracy, latency, and reliability IF THIS SOUNDS FAMILIAR You built something. It worked in a demo. But it breaks in production. That's the gap I close. Send your use case — AI, full-stack, or both. I'll tell you directly if I can help and what it takes to make it production-ready.