Iyman Ahmed / Portfolio
IymanAhmed
I ship production AI systems: agent orchestration, RAG with citations, ranking engines, legal analyzers, and financial intelligence tools.
Stockton, CA / SF Bay Area
About
Production-first AI engineering.
MSc AI from Brunel University London. Past work: 98.9% classification accuracy, production credit scoring and churn modeling, 22s → 4s latency reduction, and a legal RAG pipeline at 90% precision and 96.8% recall in production. Stockton, CA — open to SF Bay Area. US authorized, no sponsorship required.
Background
Education, experience, skills.
Education
MSc Artificial Intelligence
Brunel University London · 2022 – 2023
BSc Computer Science
Bahria University, Karachi · 2017 – 2021
Experience
Data Analyst Intern
LMP Groups · London, UK · Mar – May 2025
ML Engineer Intern
The Sparks Foundation · Karachi · Nov 2021 – Jan 2022
Key Skills
Python · TypeScript · LangChain · LlamaIndex · LangGraph · OpenAI · FastAPI · Postgres/pgvector · ChromaDB · Next.js · Docker · RAGAS · SHAP · PyTorch
Engineering Profile
What I build.
Capabilities: AI/ML engineering, LLM applications, RAG systems, agentic workflows, data pipelines
What I bring: production-grade systems, full-stack delivery, evaluation-driven development, UI integration
Stack: Python, LangChain, LangGraph, FastAPI, Postgres/pgvector, Next.js, Docker, ChromaDB
How I work: iterative, metrics-driven, deploy-first — no notebook prototypes
Featured work
Featured work.
Five production systems — outcomes, architecture, and what I owned end-to-end.
Production / Systems Credibility
Production AI systems, end-to-end.
I build full systems: ingestion, retrieval, reasoning, evaluation, and the product interface that makes it usable.
Skills / System architecture
The stack: agents, RAG, ranking, and production interfaces.
Languages: Python, TypeScript. AI/LLM: LangChain, LlamaIndex, OpenAI. Backend: FastAPI, PostgreSQL/pgvector. Frontend: Next.js, React, Tailwind, R3F. Infra/Tools: Docker, Vercel.
How I Work
Shipping beats demos.
Define a measurable target (quality, latency, cost), then build toward it with tight loops.
Design for production constraints early: observability, eval harnesses, and failure modes.
Keep interfaces premium and usable. AI is only valuable when people can trust it.
Deliver systems that survive real usage, not prototypes that only work in notebooks.
Metrics / Proof
Three systems. Six domains. Production-first.
Every project ships end-to-end: data ingestion, AI reasoning layer, and deployed interface. No notebook prototypes — only production systems.
Final core convergence
Let’s build intelligence that survives production.
Available for production AI engineering, agentic workflows, RAG systems, legal and financial AI tools, ranking engines, and premium AI product experiences.
US authorized · no sponsorship required
Contact
Start a build.
Send a note and I’ll reply with next steps.
Response time
Usually within 24 hours.