United Kingdom
I build AI systems that have to actually work—not just hit a benchmark and get shelved. Most of my projects follow the same pattern: the model is maybe 30% of the problem. The rest is deployment, failure modes, latency budgets, and the tests that catch the thing nobody thought of. I enjoy that part just as much. What I've shipped: • Multi-agent LLM pipelines on AWS at Accenture, improving analyst turnaround time by 40% for one of Indonesia's largest telecoms. • A real-time CCTV + drone dispatch system using YOLO11-seg running at 56 FPS, achieving 7.7 m 3D localisation without a depth sensor. • A clinical alerting service on Azure Kubernetes that outperformed the NHS England AKI baseline. • A self-play reinforcement learning paper on tumour localisation, published by Springer. • Currently interning at SLB, building Fourier Neural Operators and Graph Neural Networks for carbon capture simulation. MSc AI at Imperial · BEng UCL (First Class, Dean's List) · Python, PyTorch, FastAPI, Docker, Kubernetes, AWS/Azure
Stack: PyTorch, gRPC, Docker, CI/CD • Developing PyTorch-based Fourier Neural Operators (FNO) to accelerate complex fluid and thermal dynamics simulations for Carbon Capture and Storage (CCS). • Designing Graph Neural Network (GNN) architectures to map hydraulic fractures and improve spatial prediction across unseen geological layouts.
Stack: Python, FastAPI, LangChain, AWS (EC2/RDS), Docker, MCP, SQL. • Designed a unified PostgreSQL schema and API layer on Amazon RDS for 20+ finance sources, then launched an LLM-powered FastAPI + LangChain RAG service coupled with MCP on top — reducing analyst query turnaround by 40% for one of Indonesia's largest telecoms. • Reduced manual QA test-case authoring by 70% by collaborating in a 4-person team to build a GPT/Gemini multi-agent pipeline that ingested user flow diagrams and auto-generated regression test suites. • Optimised agent deployment cycle by 90% by containerising the LLM stack with Docker and standardising AWS rollout procedures, accelerating new-engineer onboarding by 20%.
• Structured fragmented data from 4+ APAC infrastructure projects to support risk and compliance decisions across $300M+ in capital investment. • Identified 10+ operational, contractual, and modelling inconsistencies across a major water supply system, improving financial model reliability for investment decisions. decisions.
• Led 6+ events, 5+ brand partnerships, 40% membership growth, 30% rise in joint participation.