Current focus
Applied AI systems
LLM products, RAG-style retrieval, structured outputs, evaluation loops, and the unglamorous engineering needed to make those systems reliable.
About
Hi, I’m Sohaib.
I’m a London-based graduate AI / machine learning engineer focused on applied AI products, LLM systems, optimization, and production-minded engineering.
I recently completed an MSc in Artificial Intelligence at King’s College London with Distinction. My work sits in the overlap between product building and technical depth: LLM workflows, retrieval systems, evaluation harnesses, FastAPI services, and optimization-backed decision tools.
Right now I’m building ContractFlow, an LLM-driven contract workflow tool, and shipping side projects that let me stress-test what modern AI coding systems are actually good at. HowlHouse is the clearest version of that so far: a spectator-first AI Werewolf platform built with Codex, then tightened through security review, CI cleanup, and multiple rounds of product polish.
Background
Optimization & decision support
MILP modelling, MiniZinc and Gurobi workflows, candidate pruning, and data-to-model pipelines that keep combinatorial systems practical.
Stack
Python-first product engineering
FastAPI, Docker, testing, structured APIs, small eval harnesses, and product surfaces that connect model behavior to actual user experience.
Links