Sarthak Gumasta
AI/ML Engineer · Generative AI | LLM | RAG Systems | Agentic System | LangChain · LangGraph · FastAPI · Azure OpenAI | Building Production AI Pipelines for Enterprise & Maritime Domains
- Role
- Ai Ml Engineer at Cyber Infrastructure (CIS)
- Location
- Indore, MP, IN
- LinkedIn followers
- 500 followers
About Sarthak Gumasta
AI/ML Engineer with 3+ years building production Generative AI, LLM, Agentic and RAG systems for maritime and enterprise clients.What I do:→ Design end-to-end AI pipelines (LangChain · LangGraph · FastAPI · Azure OpenAI)→ Build RAG systems with hybrid BM25+vector search, parent-child chunking & FlashRank reranking→ Ship defect intelligence systems processing 12-month maritime inspection data at 70% less manual effort→ Maintain 99%+ API uptime across production FastAPI services with async job dispatchKey results delivered: Cut manual analyst effort by 70–90% across multiple workflows Reduced LLM hallucinations by ~40%(verified via LangSmith faithfulness scores) Improved query resolution accuracy by 20%, response time by 25% Delivered 100% cited answers on MLC 2006 legal compliance chatbotTech: Python · LangChain · LangGraph · OpenAI · Azure OpenAI · FastAPI · Docker · ChromaDB · FAISS · Pinecone · PaddleOCR · LangSmithOpen to: Senior AI/ML Engineer · GenAI Engineer · LLM Engineer roles (Remote / Hybrid / PAN)
Experience
Ai Ml Engineer
Feb 2025 — Present
Engineered a 13-step Generative AI + RAG pipeline for maritime vessel inspection data - covering defect extraction, system-area categorisation, and sister-ship analytics - reducing manual analyst effort by 70% across production workloads.• Built a Hierarchical Parent-Child RAG system (LangChain · ChromaDB · FlashRank · Redis) with dual-path routing (Self-Querying + Hybrid BM25+Vector) delivering 100% cited answers for MLC 2006 legal compliance queries.• Applied chain-of-thought, few-shot, and structured output prompt engineering + FlashRank reranking to cut LLM hallucinations by ~40%- verified via LangSmith faithfulness scores in production.• Built a document intelligence pipeline (PaddleOCR · Table Transformer · Docling · PyMuPDF) to extract and structure tabular data from mixed PDFs - eliminating 80% of manual data-entry effort across accounting workflows.• Maintained 99%+ API uptime across production FastAPI services via exponential backoff, TPM/RPM rate-limit management, Azure Queue async dispatch, and end-to-end observability.
Education
Shri Vaishnav Vidyapeeth Vishwavidyalaya, Indore
Bachelor of Technology - BTech, Computer Science
Find verified contacts for anyone on LinkedIn
Unifers gives sales teams verified emails and direct dials, enriched profiles, and outreach that lands in the inbox.
Free plan included · No credit card required
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.