Sparsh Chawla

AI Engineer with ~3 years building production LLM applications, RAG systems, and multi-agent AI workflows. Built enterprise AI assistants handling 100–500 daily interactions using LangGraph, FastAPI/Django, Redis, and LangSmith, with working exposure to AI governance and responsible AI practices. Google Cloud Certified Professional Machine Learning Engineer with strong Python, SQL, and backend engineering experience.

Comfortable explaining AI systems and tradeoffs to technical and non-technical stakeholders, partnering across product, engineering, and business teams to move AI from proof-of-concept to reliable production use.

Data Scientist - Trine Partners - India

(2025-11)

• Built and deployed a production-grade multi-agent AI platform using LangChain, LangGraph, and Django REST APIs, with specialized search, scheduling, and calendar agents automating intent detection, entity extraction, and agent routing for enterprise event operations.
• Replaced manual database/CSV lookups with a conversational AI assistant serving 100–500 daily interactions, cutting manual processing time by roughly 50%.
• Instrumented the pipeline with LangSmith tracing and token-based cost/latency monitoring, cutting average response latency by roughly 40%.
• Built and optimized 7 reusable AI tools (participant search, arrival, departure, event lookup, date filtering, transport, meal schedules) shared across agents, applying prompt engineering to lift tool-selection accuracy by roughly 30% in the agentic workflow.
• Implemented Redis-backed conversation memory (last 5 interactions), scoped per user session under enterprise access controls, enabling multi-turn, context-aware AI assistant interactions instead of one-shot queries.
• Designed the orchestrator's response-aggregation layer to merge multi-agent outputs into a single coherent answer, and contributed to testing and deployment across the platform's full operational dataset.

Application Engineer – Data Science - Hexaview Technologies - India

(2023-08 - 2025-10)

• Built a Generative AI / RAG application for intelligent document processing (GPT-4/4o, LangChain, LangGraph) — recursive-text-splitter chunking, embeddings, and Redis-backed indexing across financial documents spanning thousands of pages — to identify stock tickers, market values, and portfolio allocations, reducing manual review time by 70%+.
• Monitored extraction and chatbot pipelines with LangSmith tracing and tiktoken token counting to track cost, latency, and output quality/error patterns as part of responsible AI practices.
• Built an HR AI assistant (Python, FastAPI, LangChain) with Redis VectorDB retrieval (RAG) and conversational memory that reached regular day-to-day adoption, answering employee policy questions without human escalation while keeping data within approved, compliant boundaries.
• Delivered an interactive AI assistant (document-analysis chatbot) producing portfolio analytics on account structure, tax classification, sector exposure, mutual-fund/retirement holdings, and stock-overlap analysis, handling sensitive financial data under enterprise access controls.
• Built and documented FastAPI services integrated with AWS API Gateway for Gate4 calculation logic, and engineered AWS Lambda/S3 ingestion pipelines automating a previously manual client workflow, applying access controls and SQL-based validation for accuracy and reliability.

B.Tech - Computer Engineering - Jamia Hamdard University (2019-06 - 2023-06)