Sarfaraz A.

Sarfaraz Ahmed – Applied AI Engineer | Agentic AI & Production GenAI I build production-grade AI systems — not demos, unstable chatbots, or prompt-only experiments. My focus is agentic AI, decision-support systems, and scalable GenAI pipelines that work reliably in real products. I help startups and teams turn GenAI ideas into stable, testable, and cost-controlled systems. I specialize in fixing the hard problems most teams face after the demo stage: • Hallucinations and inconsistent LLM behavior • Fragile RAG pipelines • Agents that fail on long-running tasks • High latency and uncontrolled API costs WHAT I BUILD Agentic AI Systems • Multi-agent systems with planning, execution, tool usage, and self-correction • Long-horizon task agents (not single-turn bots) • Supervisor, planner–executor, and recovery-based agent architectures • Reliable agent workflows using LangChain and LangGraph RAG & Knowledge Systems • Production-ready RAG pipelines over private data • Advanced chunking, metadata filtering, and hybrid search • Agentic RAG (agents that decide when and how to retrieve) • Vector databases and structured knowledge stores AI Automation & Decision Systems • Logic-driven AI tools with deterministic outputs where required • AI-powered internal tools for sales, operations, and research • Workflow automation using agents, APIs, and business rules Production Readiness • Tracing, evaluation, and observability (LangSmith-style workflows) • Cost optimization and latency reduction • Guardrails, failure handling, and safe tool execution • Multi-tenant and user-isolated AI systems RECENT WORK HIGHLIGHTS • Built a production multi-agent system (TryChirp) handling autonomous sales workflows using agent orchestration, memory, RAG, and API integrations • Designed AI prompt optimization and evaluation pipelines that reduced hallucinations and improved response reliability by over 90% • Delivered AI automation systems that replaced manual workflows and improved operational efficiency by 40–60% HOW I WORK • I separate decision logic from prompts • I design for determinism where business rules matter • I treat AI systems like software: testable, observable, and maintainable • I focus on outcomes, not buzzwords IDEAL PROJECTS • Agentic AI systems (LangChain / LangGraph) • RAG systems that must work reliably at scale • AI decision-support tools (finance, sales, operations, research) • Upgrading demo-level AI into production systems If you’re looking for someone who understands how GenAI actually fails in production — and how to fix it — I can help.