Software Engineer at SiMa.ai
I work on machine learning engineering, from model inference to vector quantization algorithms.
At SiMa.ai I work on inference for VLA (vision-language-action) models, plus agents that run locally.
My TurboQuant series shows how a random rotation makes every coordinate easy to quantize optimally.
We found MCTS a good prior to put in prompts, but in-weight approaches still beat in-context ones.
I walked through papers like DETR and GRPE, a graph transformer encoding for predicting molecule properties.
Software Engineer at SiMa.ai
I work on machine learning engineering, from model inference to vector quantization algorithms.
At SiMa.ai I work on inference for VLA (vision-language-action) models, plus agents that run locally.
My TurboQuant series shows how a random rotation makes every coordinate easy to quantize optimally.
We found MCTS a good prior to put in prompts, but in-weight approaches still beat in-context ones.
I walked through papers like DETR and GRPE, a graph transformer encoding for predicting molecule properties.
By using this service, you agree to the Terms of Service and Privacy Policy.