About me.

I'm Alpha, a machine learning engineer building on-device LLM inference, PyTorch LLM training and post-training, and GPU and AI accelerator benchmarks.

Most of my work is accelerated computing: getting models to train and run fast on whatever silicon is in front of me, whether that's an H100, a TPU pod, AWS Trainium, or a phone's NPU. My thesis, a vision-language model, was fine-tuned on Google TPUs through the TPU Research Cloud.

These days a lot of that work runs under two sibling research labs I co-founded: Experimental Machines, which benchmarks GPUs, NPUs and ASICs from the datacenter down to the phone, and Experimental Intelligence, which trains small models from scratch and labels its own data. Both publish the code, logs and raw numbers so every claim can be rerun.

What I care about beyond the benchmarks is teaching. I've given more than 20 talks, from PyTorch Conference Europe to community rooms across the Philippines, because the moment something clicks for someone in the audience is genuinely my favorite part of this field. As a GitHub Campus Expert, a lot of that has meant helping other students take their first commit, run their first model, and realize the tools are not as scary as they look.