GitHub • Projects • Coursework • Current Work
- I am a student at Imperial College London studying Economics, Finance and Data Science.
- I work through advanced ML/AI coursework and hands‑on projects.
- I enjoy building end‑to‑end data workflows: from exploratory notebooks to deployable code (CI/CD).
- I like experimenting with new tools and editors.
- Mini-GPT with Rotary Embeddings - Built a compact GPT model pretrained on the Wikipedia corpus using rotary positional embeddings, improving benchmark performance by 2.3x over standard positional embeddings. (https://github.com/slee7286/XCS224N-Work/tree/main/XCS224N%20A5/src/submission)
- Ricardo Pickle Board - Migrated a Python-based flight electronics control board backend to C++, improving runtime performance across a sensor array, 4 pyro channels, and 4 servo channels. (https://github.com/icl-rocketry/Ricardo-PickleRick)
- Kalman Filter State Estimation - Modelled a 19-state Extended Kalman Filter in Simulink fusing data from LIDAR, GPS, RTK, IMU, barometer, and magnetometer for real-time rocket state estimation, validated against a physics reference model. (https://github.com/icl-rocketry/State-Estimator-Simulink/tree/19_state_EKF)
- Flareify (ETH Oxford 2026 Hackathon) - Built a decentralised derivatives platform on Flare Coston2 supporting leveraged gas futures and stablecoin depeg protection, using Solidity contracts, a Next.js frontend, and Python/Node.js oracle infrastructure. (https://github.com/slee7286/eth-oxford-26-submission/tree/main)
- Speech-Therapy.ai (HackEurope 2026) - Built an AI-powered personalised speech and language therapy app for stroke-induced aphasia patients, using agentic AI, voice recognition, voice generation, and voice cloning to deliver accent-aware, life-relevant rehabilitation exercises. (https://github.com/slee7286/hackeurope2026)
- GlassPlate (Diamond Challenge 2024) - Built a mobile application that visualizes the environmental impact of food consumption using computer vision, barcode scanning, and location detection. (https://github.com/slee7286/GlassPlate)
- Cherokee-English Seq2Seq NMT Model - Designed a neural machine translation system with a bidirectional LSTM encoder and unidirectional decoder, achieving a BLEU score of 11.77 on the Bible corpus for an extremely low-resource language. (https://github.com/slee7286/XCS224N-Work/tree/main/XCS224N%20A4/src)
- Independent Component Analysis Audio Separation - Implemented Laplace-prior ICA to separate mixed audio signals into independent source components. (https://github.com/slee7286/XCS229-Work/tree/main/XCS229-PS5/src-ica)
- Fashion-MNIST Image Classification - Trained a neural network to classify grayscale clothing images across 10 categories from the Fashion-MNIST benchmark dataset. (https://github.com/slee7286/XCS229-Work/tree/main/XCS229-PS4/src-mnist)
- Global GDP and Youth Employment Data Analysis - Built end-to-end data pipelines in R using World Bank datasets, delivering statistical EDA and visual analytics to quantify relationships between GDP growth, youth employment, and human development. (https://github.com/slee7286/IDS_Project/blob/main/Final_Code_Deliverable.pdf)
- Spam Classifier - Implemented Naive Bayes and SVM approaches to detect SMS spam messages, comparing probabilistic and margin-based classification methods. (https://github.com/slee7286/XCS229-Work/tree/main/XCS229-PS4/src-spam)
- MuJoCo Physics Simulations - Applied REINFORCE policy gradient methods with variance reduction for continuous control tasks including hopping, running, and inverted pendulum stabilisation. (https://github.com/slee7286/XCS234-Work/tree/main/XCS234-A3/src/submission)
- Warfarin Drug Dosage Algorithms - Implemented and compared pharmacogenetic, clinical, and fixed-dose warfarin dosing strategies using LinUCB, epsilon-greedy, Thompson Sampling, and other multi-armed bandit algorithms. (https://github.com/slee7286/XCS234-Work/tree/main/XCS234-A5/src)
| Course | Description |
|---|---|
| Stanford: Machine Learning | Introduction to machine learning and statistical pattern recognition |
| Stanford: Natural Language Processing with Deep Learning | Fundamentals of natural language processing (NLP) and language models using Pytorch framework |
| Stanford: Reinforcement Learning | Main approaches and challenges in reinforcement learning |
| Imperial: Year 1 Trimester 1 | Mathematical Foundations, Probability and Statistics, Introduction to Data Science, Big Issues in Economics and Finance |
These repos show my coding work, notes (in LaTeX), implementations, and experiments.
- Advanced machine learning and optimization techniques.
- Modern NLP (transformers, attention, representation learning).
- Better software engineering practices (testing, structure, reproducibility).
- Improving my C++, Python, and Javascript skills for production‑ready apps.
- Work on redeveloping the backend of Imperial College London Rocketry's Pickle board.
