Hassan S.

Most AI freelancers build demos. I build systems that run in production — pipelines that handle real data volumes, agents that recover from failures, and architectures that your engineering team can actually maintain. My stack spans the full modern AI engineering spectrum: from stateful multi-agent orchestration with LangGraph and MCP, to production RAG systems built from scratch on Databricks, to the data engineering backbone (Spark Structured Streaming, Delta Lake, CDC) that makes all of it reliable at scale. What I Build: ~ Agentic AI Systems: Multi-agent pipelines using LangGraph — with stateful memory, conditional routing, Human-in-the-Loop approval gates, and MCP integration. Not simple chains. Systems that plan, execute, self-evaluate, and iterate. ~ RAG & Knowledge Systems: Custom retrieval pipelines (PDF parsing → chunking → embeddings → vector search) built without black-box wrappers. Adaptive routing that self-grades retrieval quality and falls back intelligently. Conversational memory across sessions. ~ Real-Time Data Pipelines: Fault-tolerant Databricks lakehouses using Medallion Architecture — Auto Loader, Spark Structured Streaming, CDC with MERGE deduplication, Unity Catalog governance, and Z-ORDER optimization across millions of records. ~ ML & Econometrics: When the problem needs statistical rigor, not just a model — causal inference, forecasting, bias/leakage checks, and calibration. Results that hold up under scrutiny, not just on a test set. Core Stack: LangGraph · LangChain · MCP · RAG Pipelines · Databricks · Apache Spark · Delta Lake · Python · FastAPI · Vector Databases · OpenAI / Anthropic APIs · MLflow · Azure · Power BI · Econometrics