Work

00Education

Yale University

B.S. Computer Science & Mathematics

Eton College

King's Scholar · Sixth Form Select · A-Levels (Math, Further Math, Physics, CS)

01Experience

Tractable

Machine Learning Intern · $1B valuation, Series E

May 2026 – PresentLondon, UK

• Built TensorRT/Triton inference pipeline for computer vision models for vehicle assessment (Mask2Former, RAFT optical flow) on NVIDIA A10G/Jetson GPUs running Linux; achieved 100% claim completions at zero error rate.
• Validated ML inference optimizations via A/B benchmarking: Valkey distributed image cache (42% latency reduction) and asyncio batch dispatch (22.1% throughput gain).

Yale Graph and Geometric Learning Lab

Student Researcher

Mar 2026 – PresentNew Haven, CT

• Researched post-training and RL for multimodal time-series LLMs for quantitative finance. Mentored by Rex Ying.
• Built LLM-based RAG pipeline synthesizing 12,216 structured market briefings across 524 S&P 500 tickers with 4-layer anti-leakage system; produced 9,843 clean training samples for fine-tuning.

SPAR Research

AI Alignment Research Fellow

Feb 2026 – PresentBerkeley, CA

• Activation steering and chain-of-thought monitorability with MIT and ex-Meta researchers.

ReefSound

Co-Founder

Jul 2023 – Aug 2025London, UK

• Built audio AI & underwater robotics achieving 99.4% accuracy classifying coral reef health from underwater sound recordings via novel CNNs and gradient boosting models; deployed across 7 countries and 3 continents.
• Featured by The Independent, NASA, National Geographic; winner of NASA Conrad Challenge against 2000 teams.
• Trained models on 130+ hours of bioacoustic recordings using spectrogram transforms and data augmentation.

Greyparrot

ML Engineer Intern · Series B, $30M raised

Jul 2024 – Sep 2024London, UK

• Optimized YOLOv8 object detection for real-time classification of 80+ industrial waste material types at production recycling facilities.
• Engineered PyTorch inference pipeline with test-time augmentation on GPU cluster for factory waste management.

CVSSP Lab, University of Surrey

Research Intern

Jun 2023 – Feb 2025Guildford, UK

• Designed novel multi-loss CycleGAN (6 custom losses: adversarial, cycle consistency, CLIP semantics, MiDaS depth, sketch extractor) for unpaired full-to-flat-color image transformation, trained on 2,246 portrait images.
• Researched graph-based audio captioning model combining BART transformer with graph convolutional network and syntax-aware dependency parsing on the Clotho benchmark.