Thomaz Bonato

About Me
Hi I’m Thomaz! I am originally from Brazil, but grew up across Sydney and Philadelphia. I have been interested in tech since I was 13, building in the EdTech space throughout high school. Outside of tech, I’m a huge fan of basketball and outdoor adventures. I enjoy reading philosophy, ethical debates, and neuroscience

What are some of your key professional/career interests?
My primary interest is on making AI actually useful—moving past demos and hype cycles to build systems that handle messy, mission critical problems in the real world. I want to build tech that endures the chaos of reality and delivers for people who aren’t engineers. The most important things to me are cutting through the noise, looking beyond benchmarks, and pushing for approaches that make a tangible, positive impact. Long-term, I’m interested in carving my own path—whether that’s launching ventures, leading research, or building teams obsessed with real-world impact.

What are some of your skills?

Applied ML, building + deploying robust AI systems, end-to-end pipelines, integrating models with real infra, production MLApplied ML, building + deploying robust AI systems, end-to-end pipelines, integrating models with real infra, production ML

What are some relevant classes/coursework?

Machine Learning, Applied Programming in Python, Statistics & Data Science , Data Structures & Algorithms, Calculus I & II, Linear Algebra for Machine Learning, Neural Networks & Deep Learning

What are some of your projects or achievements?

• ML Engineering Intern @ Netic (Founders Fund, Series A): Led end-to-end deployment of production AI systems supporting $100M+ clients and 3M+ monthly interactions. Designed scalable GCP infra for analytics, built evals for AI agents, and shipped features that accelerated our human-in-the-loop QA.
• Co-founder @ SageAI: Built a legal tech platform using LLMs for fast document retrieval and workflow automation, with 98% uptime and sub-2s latency. Secured 9 law firm partners and led ML development from zero to first $54k ARR.
• 1st Place @ HackMIT 2024: Developed a real-time ASL learning app with custom computer vision models for sign recognition. Launched as a Chrome extension with sub-second inference, earning 1st place among 1,000+ participants.
• 2nd Place @ Mercor x Etched x Cognition Hackathon: Created an AI-powered design-to-code tool that outperformed SOTA models in visual similarity and code quality. Used custom beam-search framework and CV-based LLM-as-a-judge evals to achieve production-ready results at speed.