VLVC
v.4.2 — production
After eight years operating the largest private NVIDIA H100 cluster on the U.S. East Coast (4,800 GPUs) and processing 9.4 trillion inference requests in 2024, we are still a team of 387 people solving the same problem we started with: replacing brittle in-house pipelines with infrastructure that ships, observes, and self-heals.
We didn't start VLVC to build another ML platform. We started it because the teams shipping real models to real users were spending ninety percent of their time firefighting pipelines instead of improving them. Production AI deserves the same engineering discipline we expect from databases, kernels, and compilers — nothing less, and nothing mystical on top.
Former Core ML leads from Apple, Google Brain, and Meta AI. Forty-one PhDs on staff across nine countries. Each executive below is paired with one verifiable anchor — former employer, publications, or patent count — instead of a stock headshot.
Co-founder & Chief Scientist
Former Apple Core ML lead. 19 peer-reviewed papers on production inference systems. Owns VLVC's research agenda and the Orchestrator open-source roadmap.
Anchor: ex-Apple, Core ML · 19 publications
Co-founder & CEO
Former Google Brain infra lead. Holds 7 U.S. patents in distributed training. Drives commercial strategy and the 14-months-to-9-weeks deployment methodology.
Anchor: ex-Google Brain · 7 patents
VP of Platform Engineering
Former Meta AI infra staff engineer. Owns the inference runtime, autoscaler, and the 99.992% trailing-twelve-month uptime figure across 14 global regions.
Anchor: ex-Meta AI · owns uptime SLO
VP of Research
Leads the 38-paper research output between 2020 and 2024. Principal investigator on the U.S. DOE AI Infrastructure Partner of the Year program.
Anchor: 38 papers · DOE PI
VP of Compliance & Trust
Owns the FedRAMP Moderate, HIPAA, and SOC 2 Type II audit posture for a single deployment stack — the only inference platform independently audited across all three.
Anchor: 3 frameworks · 1 stack
Chief Financial Officer
Took VLVC through $142M raised across Series A (2019), Series B (2021), and Series C (2023) from Sequoia, Andreessen Horowitz, and NEA.
Anchor: $142M raised · 3 rounds
A working record of what VLVC has shipped and what the research team has published in the last eighteen months. No embargoed pre-announcements, no roadmap theatre.
VLVC hires for technical depth rather than headcount. The platform team works in Rust and Go on the inference runtime; the systems team writes CUDA and Python against the 4,800-GPU H100 cluster; the research team contributes to the 38-paper publication record. The interview loop is a written take-home, two technical interviews, and one system-design conversation — no whiteboard riddles.
Rust · Go · CUDA · Kubernetes · eBPF
Own autoscaler and scheduling across 14 global regions. Reports to Priya Raghavan.
Rust · gRPC · Apache Arrow · FoundationDB
Build the self-healing feature store and schema-drift detection layer.
PyTorch · JAX · Triton · internal benchmarking harness
Publish 2+ peer-reviewed papers per year. Reports to Dr. Marcus Halberg.
CUDA · Triton · C++17 · NCCL · H100 / H200
Optimize the inference kernels running on the 4,800-GPU reserved cluster.
For journalists, procurement teams, and prospects running formal vendor reviews. Everything below is publicly disclosed or available under NDA on request.