Skip to content
Yoshio NomuraContact

All work · Services and data

Ignis

Maintenance intelligence for power plants

  • Built with the Pyrium team
  • Team of 3
  • October 2026
  • AMD Developer Hackathon, Act III · lablab.ai

What it is

Maintenance intelligence for power plants, built by a team of three at AMD's Developer Hackathon, Act III. Historical work orders are imported, reviewed by a person and issued as evidence-linked reports; the facts are computed deterministically and a hosted model drafts only the explanation.

Evaluation

Final valid explanation, prompt v3
21 of 24
against 14 of 24 for v2 — 12 synthetic scenarios, two runs each, DeepSeek-V4-Flash on the event's AMD API; preliminary
Model time, median, prompt v3
10.3 s
against 21.3 s for v2 — the same runs, concurrency 1
OCR mean character error, text and tables
0.000
MinerU, four synthetic document families and a held-out table image
OCR mean character error, diagrams
0.617
MinerU, the held-out synthetic diagram family

How it works

Deterministic facts and a person's review around a hosted model that drafts the explanation

Work-order exports (CSV or XLSX) are imported with explicit column mapping, and a person accepts or corrects the suggested annotations. Exact facts and priority alerts are computed deterministically, and an evidence pack scoped to what applies is handed to a hosted model on AMD's inference API, which drafts the explanation. A person approves the draft before it is issued as a PDF that links each claim to its evidence. Documents are read natively or through MinerU OCR, and runs persist in SQLite. Yoshio developed the work-order pipeline, the hosted-inference layer, document ingestion and the benchmark harness behind every figure on this page; the frontend and deployment are teammates' work.