About
unLAP is a company built around one stubborn infrastructure problem.
When expensive queries inevitably slip into production, they trigger a chaotic chain reaction of manual triage, cross-team friction, and engineering attention. This unnecessary hurdle eats into your team's velocity, taking time away from shipping new features.
We built unLAP to replace that manual toil with a high-confidence, autonomous pipeline. By dropping in a single import, our engine handles the heavy lifting of observing workloads, generating mathematically verified candidates, and validating them against regression tests, so the final rollout is a simple, low-risk decision rather than a month-long research project.
Who’s building unLAP
Meet the founders:
We are two brothers passionate about building technical solutions for large-scale optimization problems - optimizing both digital infrastructure and team productivity.
Gavin Lo
Co-founder & CEO
Before unLAP, Gavin worked as a Machine Learning Engineer at Palo Alto Networks, building AI agents, constrained LLM systems, and custom retrieval pipelines for live internal analytics.
More about Gavin
Before that, Gavin worked in security engineering at Praetorian and NVIDIA, where he uncovered critical vulnerabilities for Fortune 500 companies and built custom utilities to evade on-disk endpoint security. He also published research at Georgia Tech on automated statistical debugging, building tooling that detected 42% of the test cases missed by state-of-the-art academic suites.
His background spans machine learning, distributed systems, automated debugging, offensive security, and infrastructure-heavy systems design. He has contributed to major open source projects including Hugging Face and rclone, and has placed first in major security CTFs, including MITRE’s national cybersecurity CTF and multiple Lockheed Martin Grand Cyber Challenge events.
Justin Lo
Co-founder & CTO
Justin left Georgia Tech to work on unLAP. Before that, he worked on open-source AI systems and emerging model tooling, including contributions to AUTOMATIC1111 and other generative systems.
More about Justin
Justin has also built various other open-source ML projects, such as the autoMBW automerger and its sibling SD-Silicon model, which together earned more than 500 GitHub stars. autoMBW optimized days of human-powered merging into hours of machine time while significantly improving final output quality.
Before unLAP, Justin spent years in Linux administration, compiler development, and infrastructure engineering.
He also ran his own computer sales business.