DeepSeek V4 Flash 0731 on 8x NVIDIA DGX Spark: deployment, reproducible benchmarks, optimization profiles, and sanitized evidence.
-
Updated
Aug 3, 2026 - Python
DeepSeek V4 Flash 0731 on 8x NVIDIA DGX Spark: deployment, reproducible benchmarks, optimization profiles, and sanitized evidence.
Open-source, evidence-first financial research platform. Local Llama 3.1 70B via Ollama; signals trace to verifiable SEC filings and are hash-anchored in a public verification ledger. Research and education only — not investment advice. MIT.
A Kubernetes-native observability stack for detecting 'Silent Packet Loss' in AI training fabrics. Simulates NVIDIA DGX-style multi-interface networking using KinD, Multus CNI, and a custom Python telemetry agent with Prometheus/Grafana integration.
Production-ready local AI agent stack for NVIDIA DGX Spark / ASUS GB10. Gemma 4 31B NVFP4 + bge-m3 embeddings + OpenClaw gateway, one docker compose up.
ARM64 NVIDIA GPU evidence lab for official Folding@home contribution work
Reference C++ FIX 4.4 client implementation over standard TCP. ARM64 compatible. Not used in production signal flow — see yuclaw-brain.
Research scheduler using the Chinese Remainder Theorem for concurrent financial tick processing. Includes arXiv-style paper. Not used in production signal flow.
vLLM serving runner + LLM evaluation testplan for NVIDIA DGX Spark (GB10)
Add a description, image, and links to the nvidia-dgx topic page so that developers can more easily learn about it.
To associate your repository with the nvidia-dgx topic, visit your repo's landing page and select "manage topics."