Pavel Stoev M.Eng. P.Eng.

Pavel Stoev M.Eng. P.Eng.⁠

About

I build self-improving AI agent systems at NVIDIA, and I have worked on AI acceleration and infrastructure security.

Experiences

  1. Engineer at NVIDIABuilding agent loops that watch themselves

    My task agents implement, critique and verify, while an overseer meta-agent watches for drift and weak tests.

  2. Engineer and CISO and TPM and Partnerships at CentMLWore engineer, CISO, TPM and partner hats

    At CentML we built a fully fused GPU kernel for an entire MoE layer, speeding computations up 2 to 11x.

  3. Side projectMade Codex and Claude pair-program

    One agent implements, the other critiques and verifies against my PRD. It loops until both agree it's done.

  4. Home labRan DeepSeek V4 on two clustered DGX Sparks

    Over an 8-hour agentic run at home I measured a 95.8% prefix-cache hit ratio. In AI, cache is king.

  5. vLLM meetupHosted a vLLM meetup in Toronto

    I hosted a meetup and open discussion on vLLM, the inference engine, on a Thursday in September.

Follow Pavel for more...

XLinkedInWebsite

© 2026 Delphi · Terms · Privacy

Pavel Stoev M.Eng. P.Eng.

Pavel Stoev M.Eng. P.Eng.⁠

About

I build self-improving AI agent systems at NVIDIA, and I have worked on AI acceleration and infrastructure security.

Experiences

  1. Engineer at NVIDIABuilding agent loops that watch themselves

    My task agents implement, critique and verify, while an overseer meta-agent watches for drift and weak tests.

  2. Engineer and CISO and TPM and Partnerships at CentMLWore engineer, CISO, TPM and partner hats

    At CentML we built a fully fused GPU kernel for an entire MoE layer, speeding computations up 2 to 11x.

  3. Side projectMade Codex and Claude pair-program

    One agent implements, the other critiques and verifies against my PRD. It loops until both agree it's done.

  4. Home labRan DeepSeek V4 on two clustered DGX Sparks

    Over an 8-hour agentic run at home I measured a 95.8% prefix-cache hit ratio. In AI, cache is king.

  5. vLLM meetupHosted a vLLM meetup in Toronto

    I hosted a meetup and open discussion on vLLM, the inference engine, on a Thursday in September.

Follow Pavel for more...

XLinkedInWebsite

© 2026 Delphi · Terms · Privacy

By using this service, you agree to the Terms of Service and Privacy Policy.