AMD Panel

Token Factory · Gemma-4-31B-IT · Evolus Business Agent calls

Primary Model

Gemma-4-31B-IT

via AMD Token Factory

Evolus API Calls

20

extractor · agent · ai-tools

AMD GPU Roadmap

PyTorch / ROCm

LSTM blockage classifier

Pipeline Stats

Token Factory Calls

0

Evolus Calls

20

Avg Evolus Latency

4531ms

Evolus Calls (last 50)

Time Event Model Latency Shipment
17:29:11 compliance_notes_generated — 3806ms 2e0e06c6
17:29:06 finance_digest_generated — 3718ms 2e0e06c6
17:29:01 provider_outreach_draft_generated — 3462ms 2e0e06c6
17:28:56 provider_outreach_draft_generated — 4269ms 2e0e06c6
13:21:11 compliance_notes_generated — 3451ms 55d341c8
13:21:06 finance_digest_generated — 3731ms 55d341c8
13:21:01 provider_outreach_draft_generated — 4313ms 55d341c8
13:20:55 provider_outreach_draft_generated — 3877ms 55d341c8
13:46:24 compliance_notes_generated — 6565ms 08e75430
13:46:16 finance_digest_generated — 6263ms 08e75430
13:46:08 provider_outreach_draft_generated — 7904ms 08e75430
13:45:59 provider_outreach_draft_generated — 7713ms 08e75430
08:56:10 compliance_notes_generated — 4299ms 7a36a62c
08:56:05 finance_digest_generated — 3138ms 7a36a62c
08:56:00 provider_outreach_draft_generated — 3558ms 7a36a62c
08:55:55 provider_outreach_draft_generated — 3937ms 7a36a62c
07:34:12 compliance_notes_generated — 5081ms 6e8af23a
07:34:05 finance_digest_generated — 4388ms 6e8af23a
07:34:00 provider_outreach_draft_generated — 3547ms 6e8af23a
07:33:55 provider_outreach_draft_generated — 3613ms 6e8af23a

AMD Compute Architecture

Inference — AMD Token Factory

  • • Gemma-4-31B-IT (primary)
  • • Qwen3.8-Flash-Next (fallback 1)
  • • DeepSeek-V4-Flash (fallback 2)
  • • Semaphore: ≤4 concurrent · ~15 RPM

Training Roadmap — AMD ROCm GPU

  • • LSTM blockage classifier (PyTorch)
  • • GDELT + PortWatch training data
  • • Run: make train
  • • Artifact: models/lstm_blockage.pt