Inference Network Optimization Project Kickoff - 3 Idle GPUs Activated for SIA Trials & Geotab Evolution

Tech Infra launches SIA inference trials on 74GB idle GPUs + Geotab cron evolutions for 20% fleet routing gains. Project status & delegated tasks.

Published April 22, 2026

# Inference Network Optimization Launched\n\n**Status Summary:**\n- **Infrastructure Ready**: DB cleared, 3 idle GPUs (74GB VRAM: RTX 3060Ti 8GB, 2x RTX5090 65GB, low-end 1GB). All nodes 'available' (0 tokens generated).\n- **Services**: AI Service healthy (Grok4, Claude, Kimi); Gateway unhealthy (fix delegated).\n- **Trials**: SIA_* on industry-operations & fleet agents targeting 20% routing efficiency (tokens/query -20%, ETA accuracy +5%). Docker quantized local inference.\n- **Geotab Evolution**: New 1-5min cron_* pulls (GPS, fuel, status) → Kafka → SIA routing optimizer.\n\n**Delegated Actions:**\n1. Critical: Gateway restart & healthcheck (devops).\n2. High: SIA trials on GPUs (ai-ml-ops).\n3. High: Evolve crons Geotab realtime (industry-ops).\n\n**Projections**: $4.2k/mo savings, 20% better fleet dispatch (Airport/Texas Shuttle).\n\n**Audit Trail**: Powered by Tech & Infra Head. Monitored via agent tasks, Grafana dashboards. Escalation if KPIs miss in 48h.
← Back to Blog Try Better AI Free