Speaker
Description
KubeEdge Ianvs is a CNCF distributed AI benchmarking toolkit used to test machine learning workloads across edge computing environments. As an LFX Mentee with CNCF, I spent four months working through a problem that doesn't get talked about enough in AI tooling: benchmarks that work on one machine and silently break on another.
The core issues weren't algorithmic — they were dependency drift across distributed AI examples, Python version inconsistencies that only showed up across different environments, and runtime configurations that worked by accident rather than by design. Fixing them meant tracing failures back through dependency chains, reproducing environment-specific bugs that didn't show up locally, and rebuilding configurations so they'd actually hold up across setups instead of just working on the maintainer's machine.
This talk covers what those bugs actually looked like, how I diagnosed them without access to every environment they failed in, and what it taught me about writing benchmarking code that's reliable across machines — not just correct on one. I'll also talk about what it's like mentoring under CNCF: how LFX mentorship actually works, what maintainers expect from contributors, and how to get unstuck when a bug only reproduces in someone else's environment.
No prior Kubernetes or CNCF background required, though basic familiarity with distributed systems concepts will help.
What you'll leave with:
1.A real example of environment-specific bugs in distributed AI tooling and how to track them down
2.Practical lessons on writing benchmark code that survives different setups, not just your own
3.A picture of how CNCF/LFX mentorships work, for students considering applying
Any other info we should know?
This talk is technical, grounded in completed LFX mentorship work (Aug–Dec 2025) with CNCF on the KubeEdge Ianvs project — a real distributed AI benchmarking toolkit, not a hypothetical or toy example. It assumes basic familiarity with distributed systems concepts; no deep Kubernetes expertise required.
The audience will get concrete debugging strategies for environment-specific failures (dependency drift, cross-environment inconsistencies, configs that work by accident), plus a realistic picture of how CNCF/LFX mentorship programs actually work — useful for students considering applying to one.
The session will be interactive, with Q&A taken throughout rather than saved for the end, since attendees are likely to have specific follow-up questions on debugging approach and tooling choices as they come up.
Session author's bio
Abhishek Kumar is a Computer Science graduate from Dr. A.P.J. Abdul Kalam Technical University, currently a GSoC 2026 contributor at the United Nations Office of Information and Communications Technology (UN-OICT), building a production AI pipeline for FireForm — selected from over 200 applicants for one of two spots. He previously worked as an LFX Mentee with the Cloud Native Computing Foundation (CNCF) on the KubeEdge Ianvs benchmarking toolkit, and as a Software Development Engineer Intern at Akatsuki in Japan, building a production AI-powered translation platform. He's an ICPC Asia West Regionalist .
| Level of Difficulty | Beginner |
|---|---|
| Please confirm that there are included headshots of all speakers in their profiles | Yes |
| Agree to Privacy Policy and Notice | I agree |
| In Person Attendance | In-person |
| Social Media | linkedin.com/in/abhishekkumarji |