Vivek Suryavanshi

Senior GenAI Engineer · Agentic AI Architect

I build AI that acts — reliably.

14+ years in enterprise software. 4+ years taking agentic AI from POC to production — LangGraph, RAG pipelines, and rigorous LLM evals. 4x Anthropic Claude Certified.

View Projects Get in Touch

14+Years in software

4+Years in agentic AI

4xClaude Certified

10k+Users served in production

meet_the_agent.pseudo

architect_agent = Agent(role="GenAI Architect")

while architect_agent.is_running:
    think()    # plan the steps
    act()      # call tools, fetch data
    verify()   # trust, but check

>>> shipped to production ✓
>>> faithfulness: 97% · confidence: 94%

About

My runtime Philosophy.

> architect_mindset

Enterprise RAG platforms on Azure serving 500+ employees — designed for scale, not demos.

> eval_obsessed

Multi-level eval design — end-to-end, trajectory, and component — with golden datasets and LLM-as-judge scoring.

> builder_not_theorist

Four live agentic apps shipped solo — voice RAG, multi-agent planners, and natural-language data tools.

Senior GenAI engineer and Solutions builder who takes AI systems from proof-of-concept to production — shipping Agentic AI, RAG, and LLM systems along the way. My work spans designing agentic workflows with LangGraph and agent SDKs, integrating enterprise tools and data, and building rigorous evaluation pipelines across every layer of an agent — so quality isn't a guess, it's measured. I'm Anthropic Claude-certified across both the Architect and Developer tracks, and I spend as much time architecting and evaluating agents as I do in stakeholder rooms shaping AI roadmaps.

Skills

The Agentic AI stack, end to end.

Agentic AI & Orchestration

LangGraph, Agentic RAG, OpenAI Agents SDK, LangChain, CrewAI, Planner–Reasoner / ReAct, Memory & State Mgmt, Tool Orchestration, Guardrails, Prompt Engineering (CoT, few-shot)

Agentic Dev & Tooling

Claude Code, MCP Servers, Agent Harness Design, Spec-Driven Dev

Agentic & LLM Evaluation

Multi-Level Eval Design, LLM-as-Judge, Golden Datasets, Ragas, Tool-Selection Benchmarking

LLM Platforms & Models

Anthropic Claude, Azure OpenAI, OpenAI (GPT-4o, Whisper, TTS), Google Gemini / Vertex AI, Hugging Face, Cohere, Ollama / MLX

RAG & Retrieval

Hybrid Retrieval, Azure AI Search, Chunking & Embeddings, FAISS, Pinecone

LLMOps & Observability

Langfuse, Tracing, Cost & Latency Monitoring

Cloud & Backend

FastAPI, Python, Azure OpenAI, Azure AI Search, Cosmos DB, GCP / Vertex AI, Pydantic, Streamlit, Docker, Jenkins CI/CD

Delivery & Leadership

AI Roadmap Planning, Stakeholder Alignment, Technical Discovery, Workshop Facilitation

Experience

Fourteen years, one steady climb towards Autonomy.

Senior GenAI/LLM Engineer & Solutions Architect

Mar 2022 – Present

Accenture Solutions Pvt Ltd, Gurugram

Architected end-to-end enterprise RAG platforms on Azure (Azure OpenAI, AI Search, Cosmos DB, Blob Storage) — serving 500+ employees across the US with policy-grounded answers.
Designed agentic workflows with LangGraph & LangChain, applying planner–reasoner and ReAct patterns to orchestrate retrieval, tool execution, and state management.
Built RAG evaluation & observability pipelines — LLM-as-judge scoring for faithfulness and retrieval accuracy, with Langfuse tracing for iterative production quality.
Defined agent evaluation metrics (tool-selection precision, parameter accuracy, trajectory coverage) using golden datasets.
Optimized chunking, embeddings, and hybrid retrieval (BM25 + vector) — cutting token consumption ~25% while improving retrieval accuracy.
Built FastAPI inference APIs with structured Pydantic outputs and auth for production-grade GenAI services.
Configured MCP servers (PostgreSQL, Playwright) in VS Code — enabling QA to run agentic test workflows via GitHub Copilot agent mode.
Led 10+ technical discovery sessions, translating business problems into GenAI solution blueprints; co-owned the roadmap across 3+ internal AI assistant initiatives.

Business Analyst — Digital & AI Solutions

Mar 2021 – Mar 2022

Accenture Solutions Pvt Ltd, Gurugram

Translated business requirements into functional specs, KPIs, and success metrics for AI-enabled workflows; ran stakeholder workshops and UAT cycles driving measurable adoption.

NLP Engineer (Python)

Mar 2019 – Feb 2021

Accenture Solutions Pvt Ltd, Gurugram

Built Python NLP preprocessing pipelines (spaCy, NLTK) for cleaning, NER, tokenization, and lemmatization — the foundation for later LLM work.

Senior Automation QA Engineer

Apr 2011 – Feb 2019

Accenture Solutions Pvt Ltd, Gurugram

Led test automation (Selenium, TestNG, Java, SQL) with 75%+ coverage; integrated Jenkins CI/CD, reducing regression effort by ~40%.

Projects

Shipped, not just prototyped.

🎙️

Live App GitHub

WhisperRAG

Voice-Powered RAG with Production Observability

Voice RAG grounded strictly in a loaded webpage. FastAPI + OpenAI Whisper / GPT-4o-mini / TTS, Supabase session logging. Langfuse tracks cost, latency, and traces; LLM-as-judge runs faithfulness scoring on every conversation.

FastAPIWhisperLangfuseLLM-as-Judge

🧩

Live App GitHub