Ananta P.

Most AI projects die between the demo and production. I close that gap: LLM agents, RAG, text-to-SQL, and the data pipelines and warehouses underneath them. 70 contracts, 1000+ hours, Top Rated. Working in data since 2012. MTech in Data Science from BITS Pilani. 8 years at Accenture, 4 at Persistent Systems. Most AI projects stall in the same place. The demo works, then it meets real data. That gap is where I spend my time: ingestion, modelling, evaluation, and the unglamorous reliability work that decides whether an agent survives contact with production. WHAT I BUILD AI and LLM systems - LLM agents and multi-agent workflows, with tool use, retrieval and evaluation harnesses - Text-to-SQL over real warehouses, with guardrails so business users can self-serve - RAG pipelines: chunking, embeddings, retrieval quality, hallucination control - AI features embedded into existing products, not standalone demos Data engineering - ETL and ELT pipelines from source systems into the warehouse - Data quality, validation and pipeline observability - Warehouse and lakehouse modelling - Platform migrations and modernization Analytics people actually use - Apache Superset and Preset, across multiple long-running engagements including 100+ hour front-end builds - Executive dashboards, embedded analytics, custom visualizations - Reporting layers that keep working after handover RECENT WORK - Text-to-SQL interface that lets executives query the warehouse in plain English - ESG data ingestion and automation for compliance reporting - Predictive maintenance models for manufacturing - Superset front-end and visualization work across several repeat clients STACK Python, SQL, Claude and OpenAI APIs, LangChain, FastAPI, PostgreSQL, MySQL, Apache Superset, Apache NiFi, Airflow, Elasticsearch, Power BI, AWS, Docker HOW I WORK Small scope first. I would rather prove the approach on a paid discovery or one narrow slice than write a long proposal for something neither of us has tested. You get working code, documentation and a real handover, not a dependency on me. Good fit if you have real data, a real business problem, and you want something running rather than a prototype. Not a fit if you need someone to fill a seat and work a ticket queue. Tell me what you are building and where it is stuck. I will tell you honestly whether I am the right person for it.