Currently building at 
Research
ResearchShay Manor — AI/ML Engineer
Building cool things at scale
I create scalable machine learning systems, train large language models, and explore quantum-enhanced algorithms to solve real problems.
ResearchI create scalable machine learning systems, train large language models, and explore quantum-enhanced algorithms to solve real problems.
Fine-tuned Llama 3.1 8B to 48% MFU and +28% MT-Bench, and shipped an automated fine-tuning platform with 1k+ image pulls and load-balanced vLLM inference.
Teacher-guided failure detection for distilled semantic segmentation under domain shift, and physics-informed horizon reduction for offline reinforcement learning.
Leading ~30 engineers building a self-driving stack from scratch — 1st place at the International Autonomous Karting Series as a rookie team.
Team lead of Purdue Electric Vehicle's Autonomous team (~30 people). Responsible for fully designing and implementing self-driving solution from scratch. Using first-ever 3DGS pipeline for autonomous racing on-device.
A custom 1.3B parameter GPT2-style language model pretrained on 500B tokens from diverse sources including Oscar, OpenWebText, Wikipedia, and ArXiv, achieving impressive perplexity scores.