Ankit S.Private profile: Upwork shows its earnings only to signed-in members, so we check it every few weeks, not daily. MRR here is a rough estimate (Upwork's six-month earnings ÷ months active). Last read Nov 5, 2026.

I build AI agents that actually run in production, not demos that break on the second query. If you need an LLM-powered system that is reliable, cost-efficient, and maintainable, that is exactly what I do. Most "AI developers" have built chatbot wrappers. I have shipped multi-agent systems at a Fortune-scale logistics company that resolve critical production incidents in under 30 minutes, systems that engineers depend on daily. WHAT I BUILD: Multi-agent systems with LangGraph: supervisor/worker patterns, stateful workflows, human-in-the-loop approval gates MCP (Model Context Protocol) servers and integrations: custom tools that let Claude/GPT securely access your databases, APIs, and internal systems RAG pipelines that stay accurate: hybrid search (vector + keyword), reranking, semantic caching, evaluation harnesses Document intelligence: OCR to structured JSON pipelines that cut LLM token costs by 60%+ vs raw image processing. Agent cost optimisation: semantic caching (I have achieved 99% cache hit rates), model routing, prompt compression. PRODUCTION RESULTS: Built an RCA (root cause analysis) agent system over 33B+ log rows (ClickHouse) + code intelligence + dependency graphs (Neo4j). Cut P1/P2 incident resolution from 4+ hours to under 30 minutes. Designed a 6-agent document processing pipeline (LangGraph + FastMCP + pgvector) with 0.92 cosine similarity caching, reaching ~99% cache hit rate and 65% token cost reduction. 10+ years of backend engineering underneath: I understand the systems your agents connect to (Kafka, PostgreSQL, Kubernetes, AWS), so integrations are solid, not duct tape. WHY IT MATTERS: The hard part of AI agents is not the LLM call. It is state management, error recovery, cost control, observability, and clean integration with your existing stack. That is distributed systems engineering, which is my core profession. TECH: LangGraph, LangChain, MCP, Claude/GPT/open models, pgvector, ClickHouse, Neo4j, Python, FastAPI, PostgreSQL, Redis, Kafka, Kubernetes, AWS, OpenTelemetry Available for: agent MVPs, RAG system builds, MCP server development, AI architecture reviews, rescuing stalled AI projects. Message me with what you are trying to automate. I will tell you honestly whether an agent is the right solution or overkill. Skills tags LangGraph, LangChain, AI Agent Development, RAG, Model Context Protocol, LLM, Prompt Engineering, Python, FastAPI, pgvector, Vector Databases, OpenAI API, Anthropic Claude, Machine Learning Integration, API Integration