Jabran K.

✅ TOP RATED PLUS — trusted to deliver on large, complex, business-critical projects. Don't take my word for it — check my reviews. ✅ 100% Job Success Score — I take ownership and see things through to completion. ✅ 25 years of professional software development — from database engineering and team leadership at S&P Global to building AI-powered portals on AWS today. ✅ 6 certifications including AWS Solutions Architect Professional and Big Data Specialty. Plus PMP and Agile Leadership. I understand architecture, delivery, and risk. If your AI works in the demo but nobody will sign it off, that is the problem I solve. I build the evaluation sets, deterministic checks, tracing and human approval gates that let a risk owner put their name on an LLM system — and I have done it inside a business where more than 700 researchers depend on the output. Most AI projects do not fail because the model is wrong. They fail because nobody can prove it is right. I spent twenty years being accountable for production systems that were not allowed to fail: 500+ job ETL estates, Level 3 escalation, options and market data at global financial-data firms. For the last four years I have applied that same discipline to generative AI as a Senior AI Application Architect — human in the loop, deterministic checks on model output, and a named approver before anything ships. WHAT I DELIVER - A failure taxonomy built from your real transcripts, not hypotheticals - A golden evaluation set with explicit pass thresholds, running in CI as a regression suite - Deterministic validators and guardrails wired into your pipeline - Tracing and alerting on the failure modes that will actually hurt you - A human approval gate placed where it catches failures without throttling throughput - A one-page memo your risk owner can sign I BUILD AGENT RIGS, NOT JUST PROMPTS Mine make engineers code faster and more reliably, and generate every artifact a company's own standards require — so people ship to the standard without having to stop and go read it. One of them is used by a 30-person data science team to ship their own production tooling. They all carry a rule that tells you how I think: the agent that builds never tests. In production I have also built MCP servers connecting custom internal applications to Claude and ChatGPT, so the people using those applications can ask what the data in front of them means instead of queuing for a developer to explain it. That removed a standing drain on engineering time. I work primarily on AWS, where I hold the Solutions Architect Professional certification, and primarily with the Claude and OpenAI APIs. TRACK RECORD - 100% Job Success Score across 19 contracts, 1,389 hours and $100K+ earned here - One of those contracts became the senior in-house AI architecture role I have held since 2022 - 11 self-serve tools shipped, turning data science scripts into applications researchers run themselves - An AI adoption dashboard used by the C-suite to measure AI usage across 700+ researchers - AWS Solutions Architect Professional, AWS Big Data Specialty, PMP WHERE I AM NOT THE RIGHT FIT I do not do model training, fine-tuning or computer vision, and I do not take on simple chatbot builds or low-level automation glue. If you need a RAG chatbot built from scratch, I am not your best option right now. HOW TO START The lowest-risk way to work with me is a five-day AI Reliability Diagnostic. You get the failure taxonomy, a first evaluation set and a prioritised risk register — a document you can take to your risk owner in week one.