A TypeScript monorepo that turned a single online board game into a platform. Rules plug into one typed interface, the server treats a match as opaque state, one set of Zod schemas is both the validator and the published API spec, and a second service runs automated tournaments between bots over WebSocket.
A teaching system built as software: a Python pipeline that generates course artifacts from source and verifies them by executing them, and a Next.js platform that authenticates learners, serves material from private storage, and runs their code either in the browser or on a pool of remote GPUs.
A computer-vision system split across a network boundary. A stateful client owns capture, presentation and storage on ordinary hardware; a stateless FastAPI service owns the model on a GPU; and everything between them — bandwidth, latency, failure and portability — is engineered at the seam.
A 12.5M-pair Vietnamese diacritic-restoration corpus, and four seq2seq architectures built from scratch in PyTorch to translate undiacritised text back into correct Vietnamese.
A systematic study of GPU inference optimisation across ResNet50 and DistilBERT — FP16, INT8 quantisation, TorchScript, torch.compile and ONNX — measured rather than assumed.
A fine-tuned YOLOv5 detector coupled with OCR, wrapped in a FastAPI service and a React frontend so citizens can look up their own traffic violations. Runs at ~12 FPS.
A quantitative trading system built from scratch in PyTorch — a Transformer classifying BUY/SELL/HOLD from technical indicators and Time2Vec embeddings over 3.4M training samples.
Fraud detection on a dataset that is 0.172% positive — where accuracy is a useless metric and the whole problem is resampling strategy and the recall/precision trade-off.
Sentiment classification on Sentiment140 with Naive Bayes and SVM, run on both Apache Hadoop and Apache Spark to measure what the execution engine actually costs you.
A ground-up PyTorch reimplementation of YOLOv1 — custom dataset, custom loss — extended with CBAM attention to measure what attention contributes to a one-stage detector.
