Engineering lead · Researcher · Founder
I co-founded and sold a healthcare software company, designing and building its hospital claims platform. At UT Austin, I develop methods that verify and improve AI decisions—helping businesses automate more work, keep errors under control, and turn expert feedback into systems that need less supervision.
Selected work · Healthcare software
I designed and built the entire claims platform. It reconciled insurer payment records against hospital bills, helping billing teams recover denied, unpaid or underpaid amounts. It resolved most incompatibilities automatically; for the rest, operators could use a resolved discrepancy as a template for the software to propose further matches, then confirm the results.
Hospitals across most regions of Brazil used it to process claims worth billions of reais. The company was acquired, the buyer's engineering team took over development, and the platform remains in production under successive owners.
Research
I first-authored Regime-Conditional Verification at UT Austin. The method estimates when a safety classifier disagrees with an operator's policy, then uses those estimates to correct decisions, detect drift and guide repairs.
Thiago Sandoval · Prof. Ufuk Topcu · 2026
I earned my computer science degree at the University of São Paulo in 2004 and am completing a master’s in artificial intelligence at the University of Texas at Austin, with graduation expected in December 2026.
Federal Forensic Expert
For thirteen years, I worked in computer forensics and cybercrime with Brazil's Federal Police. I used techniques including malware analysis, reverse engineering and data recovery to produce technical evidence for federal criminal proceedings. My work included reconstructing a banking-malware operation from its source code.
Cloud architecture
I redesigned cloud architecture for machine-learning diagnostics ahead of national rollout across a telecom network serving over 10 million subscribers.
Lead Engineer
At Stage, I lead AI innovation, working on classification and matching systems for music royalty and credit platforms serving PPL, BumaStemra and Universal Music Group. I contributed to the team’s 23% improvement in repertoire auto-match rates across 28 million recordings.
Contact
For applied AI research, product development or engineering opportunities, get in touch. Based in Cambridge, UK.