I help experienced engineers and technical founders investigate GPU performance problems and make better decisions about inference systems. My focus is auditing GPU usage and workload patterns, investigating inference bottlenecks, and reviewing low-level system design. Bring a workload you want to understand, a performance problem you’re investigating, or an architecture decision you need to make. We’ll examine the available evidence, discuss trade-offs and identify practical next steps. I’ve spoken at EuroPython about writing fast GPU code in Python, and publish technical articles and interactive explainers at abhik.ai. I keep pro bono conversations for students, open-source contributors and people who cannot comfortably afford a paid session. You don’t need to explain your finances. If you’re booking for a business or employer, please choose a paid session.