Model improvement platform
Crucible
A platform for building evals, fine-tuning open-source models, comparing candidates, and tracking reproducible model lineage across local, SSH, and Slurm compute.
Live appRutgers CS + Data Science
Alejandro Pinto
I build machine learning systems, developer tools, and high-performance engines, with recent work across Amazon, Rutgers research, Harvey Mudd ML research, and autonomous systems.
Rutgers CS + Data Science, 3.95 GPA
Chexagon Hexagonal chess engine reaching 500K-900K nodes/sec
AWS Outposts console dashboard with 90%+ test coverage
About
Focused on systems that make ML work easier to run, inspect, and trust.
I am studying computer science and data science at Rutgers, with a mathematics minor and coursework spanning graduate AI, regression methods, linear optimization, and data structures.
My recent work includes a React and TypeScript dashboard for the AWS Outposts Console, learnable steering methods for MusicGen, Crucible for model-improvement workflows, and a Go engine for hexagonal chess.
Engine + WASM demo
Chexagon
An undefeated 91-square hexagonal chess engine in Go with alpha-beta search, transposition tables, quiescence search, and NNUE training work.
Play demoML research
MusicGen steering
A contrasting concept pair generator and compute-efficient steering pipeline validated with CLAP score, Frechet Audio Distance, perplexity, and PCA plots.