ARI XPands / bowser
Bowser is my 26-week AI Systems Mastery build, focused on learning to research, train, optimize, deploy, and operate modern AI systems end to end. It spans computer vision, multimodal agents, LLMs, efficient training and inference, ML platforms, distributed systems, systems programming, MLOps, evaluation, and reliability, brought together through four integrated capstone systems with reproducible experiments, benchmarks, tests, and public technical evidence...
Built and tested a simple PyTorch FashionMNIST classifier end to end. Implemented data loading, model definition, training, validation, test evaluation, checkpoint saving/loading, and a FastAPI inference interface. Final test accuracy reached 87.74%, with 5 automated tests passing successfully.
Set up Bowser as the foundation for the 26-week Xpand with uv, pytest, ruff, and mypy, verified the environment with a smoke test, and established the initial project structure. Built and tested a concurrent log analyzer over 2 million synthetic log lines, benchmarked different worker counts, profiled the implementation to understand performance bottlenecks, and completed a Top K DSA problem as part of the Week 1 baseline work.