Post Job Free
Sign in

GenAI & LLM Multi-Agent Engineer

Location:
Surat, Gujarat, India
Salary:
8 LPA
Posted:
October 30, 2026

Contact this candidate

Resume:

YUVRAJ VARMA

GenAI & LLM Engineer Multi-Agent Systems Applied AI Engineering

Surat, Gujarat, India +91-903******* ******.*******@*****.*** LinkedIn GitHub Portfolio PROFESSIONAL SUMMARY

GenAI/LLM Engineer with 2.5+ years designing and shipping production-grade Large Language Model (LLM), Retrieval-Augmented Generation (RAG), and multi-agent AI systems. Hands-on experience with LangChain, LangGraph, OpenAI GPT, Azure OpenAI, and Google Gemini across text, vision, and speech modalities. Skilled at building FastAPI backends and deploying applied AI systems on Microsoft Azure end to end, from prototype to production. Experienced translating ambiguous business problems into structured, agent-driven AI workflows. TECHNICAL SKILLS

● LLMs & GenAI: GPT-4/GPT-4o (OpenAI, Azure OpenAI), Google Gemini, Anthropic Claude, Groq, Mistral, Perplexity, LangChain, LangGraph, Retrieval-Augmented Generation (RAG), Multi-Agent Systems (CrewAI), Prompt Engineering, Model Context Protocol (MCP)

● Computer Vision & Diffusion: Stable Diffusion, OOTDiffusion, VITON-HD, TryOnDiffusion, ControlNet, Image-to-Image Generation

● Speech & Multimodal AI: Azure Speech-to-Text/Text-to-Speech, OpenAI Whisper, GPT-4o Translation, Rhubarb Lip Sync

● Backend & Cloud: Python, FastAPI, Microsoft Azure (Azure OpenAI, Cognitive Services), REST APIs, Make.com Workflow Automation

● ML & Data Engineering: TensorFlow, Keras, Scikit-learn, NLP, Time Series Forecasting (Facebook Prophet), Pandas, NumPy, SQL, R, Apache Kafka, Apache Airflow

WORK EXPERIENCE

AI/ML Engineer — Narola AI Studio, Surat, Gujarat

December 2024 – Present

● Designed and deployed a multi-agent receipt-analysis pipeline using Vision AI and GPT for document understanding, automating structured data extraction across varied receipt formats and reducing manual data-entry effort.

● Built a full-stack financial content platform integrating Google Gemini (with grounding) and the SEC XBRL API, generating structured, fact-grounded risk and market-positioning reports from real-time company financial data via a FastAPI backend.

● Engineered a generative visual pipeline (OOTDiffusion) converting flat-lay product photography into lifestyle imagery, outperforming VITON-HD and TryOnDiffusion benchmarks by 90–95% on realism.

● Built an interactive real-time 3D avatar (Ready Player Me, Mixamo, Rhubarb Lip Sync) with OpenAI-driven conversational voice interaction for low-latency customer engagement.

● Developed a bidirectional, multilingual speech-translation system combining Azure STT/TTS, GPT-4o translation, and LangGraph agent orchestration for real-time listen-translate-respond conversations.

Research Apprentice — Jain (Deemed-to-be) University, Bengaluru, Karnataka February 2023 – September 2023

● Forecasted UPI (Unified Payments Interface) transaction growth using time-series analysis and Facebook Prophet, projecting a 66% increase in monthly transaction volume to inform stakeholder planning around India's digital payments infrastructure. PROJECTS

F1 Strategy Simulator Multi-Agent AI, CrewAI, FastAPI, React Built an AI-powered race-strategy simulator using a CrewAI multi-agent architecture (data analyst, weather specialist, race strategist agents), FastF1 telemetry, and OpenWeatherMap data; pre-computed tire-wear physics per compound and generated driver-specific, lap-by-lap pit strategies via Groq-hosted LLMs, deployed with a FastAPI backend and React/TypeScript frontend.

Second Opinion — AI Document Analyst FastAPI, LangChain, RAG, Gemini Built a multi-stage document-analysis pipeline using Docling for universal format parsing (PDF, DOCX, PPTX, XLSX), RAG-based retrieval over document content, and Gemini for automatic domain detection (legal, medical, financial, technical), surfacing plain-English summaries, red-flag clauses, and an overall risk verdict. Rock-Paper-Scissors Plus: AI Judge LangChain, Gemini, Pydantic Designed a prompt-driven decision engine where an LLM classifies free-text game moves — handling typos, synonyms, and ambiguous input — via a three-layer intent-logic-response architecture with structured Pydantic outputs. EDUCATION

MSc. Data Science & Analytics CGPA: 8.99/10

Jain (Deemed-to-be) University, Bengaluru, Karnataka 2022 – 2024 Bachelor of Computer Application CGPA: 8.15/10

SDJ International College, Surat, Gujarat 2019 – 2022 CERTIFICATIONS & ADDITIONAL ACTIVITIES

● Certifications: Prompt Engineering for Generative AI (2024); Data Analysis with Alteryx, LinkedIn Learning (2024); Data Manipulation & Analysis with Pandas/NumPy, LinkedIn Learning (2023)

● Content Head, Datum (Jain University) — initiated and managed a monthly newsletter covering Data Science and AI advancements for the university community

● Freelance Commerce Tutor (Feb 2020 – Jun 2022) — supported 12th-grade students, achieving an 80% success rate in grade improvement



Contact this candidate