Architect · AI on GPUs · Digital-asset systems

I build systems where accuracy matters.

AI for regulated industries: self-hosted LLMs on H100, L40 and L40S, and document AI for financial and regulatory filings. Trading infrastructure: market-making bots in Go, a trading platform in Rust, smart contracts in Solidity. I keep measured results and estimates clearly apart.

  • 3GPU types in production: H100 · L40 · L40S
  • 3languages for trading and contracts: Go · Rust · Solidity
  • 4models calibrated on real vLLM runs
~/gpu-fit
$ pip install gpu-fit
$ gpu-fit Qwen/Qwen3-8B --gpu L40S --context 32k --concurrency 4

Qwen/Qwen3-8B  on  1× L40S (45.0 GiB)
  vLLM budget per GPU          41.4 GiB
    weights                    15.3 GiB
    runtime overhead            1.9 GiB
    left for KV cache          24.3 GiB
  KV cache capacity       176,649 tokens
  Requests at full context        5.4

✓ Fits. Suggested command:
  vllm serve Qwen/Qwen3-8B --max-model-len 32768
Before you rent the GPU: try gpu-fit in your browser

Where I work

One thread: technology that has to be right.

Life sciences

Enterprise architecture and technology leadership in global pharma, where validation and change traceability are central.

Tax & finance

AI platforms for transfer pricing and tax compliance: reading financial statements, benchmarking, drafting reports used for real filings.

AI in production

Self-hosted LLMs with vLLM on H100, L40 and L40S: AWQ and FP8, shared-GPU budgeting, admission control and self-healing. Plus long-document AI and gateways with per-call cost tracking.

Digital assets

Market-making bots in Go, a trading platform in Rust, smart contracts in Solidity (EVM, Arbitrum) and Solana programs in Rust. Latency, keys, ledgers and edge cases first.

How I work

  1. 01

    Measure, then claim.

    Every number I publish comes from a run on real hardware or arithmetic you can check. Estimates are labelled.

  2. 02

    Deterministic where it counts.

    Money, ranges and percentiles are code with tests. AI does the reading and drafting, never the maths.

  3. 03

    Private by design.

    When data can’t leave, the model comes to the data: open models on your own GPUs, sized for the workload.

  4. 04

    Stay hands-on.

    I lead teams and architecture, and I stay close to the code, the benchmarks and the logs.

Open source

gpu-fit

How many users can this LLM serve on this GPU? A parameters-times-bytes estimate is a good start; gpu-fit goes further. It reads the model’s real architecture (grouped-query, hybrid, MLA, sliding-window, FP8) and mirrors vLLM’s memory budget. Then it hands you the command.

  • Measured on three GPUs across Ampere and Ada (L40S, A10G, L4); H100 next
  • Shared GPUs: --reserve-gib suggests the memory setting
  • Warns before vLLM refuses to start a hybrid model
  • No dependencies · MIT licence · Python 3.10+
Try it in the browser →

Also in the open: gpu-guard (Go, in development), a health check that flags a model whose speed drops, a common sign it has fallen back to the CPU.

Grouped-query attentionHybrid linear attentionDeepSeek MLASliding windowFP8 / AWQ / GPTQTensor parallelFP8 KV cacheVision encodersShared GPUs

Writing

Field notes

Building something where accuracy isn’t optional?

I’m always glad to compare notes on private LLMs, document AI, trading infrastructure or smart contracts.

Find me on GitHub