ContactAveriom
The operating system for physical safety: learn tacit constraints from human micro-corrections, predict boundary violations, and apply tiered inhibition — with evidence by design.
Why now
Automation is too rigid; manual work is too risky. Physical AI is moving fast, but safety and trust trails capability. Averiom is the missing control layer: learn negative constraints, predict violations, and enforce envelopes at the edge.
Moat
- Constraint learning from real correction events (not scripted labels).
- Edge-native runtime (latency budget is designed, not hoped for).
- Portable profiles + evidence events (auditability baked in).
- Cross-domain primitives: the same stack maps to many verticals.
Raise & milestones
SOURCE: HCII2026 · Springer CCIS · “Inverse Constraint Learning for Real-Time Manufacturing Safety” · Tushar Mishra, Averiom (2026)
• Engineering: finalise v1.0 edge runtime (target latency < 5ms)
• Pilot: design-partner deployment in composites welding or autonomous mining
• IP: convert provisional → non-provisional (deadline Jan 2027)
• Academic: HCII2026 presentation and Springer publication (Jul 2026)
- Design-partner agreement in composites welding or mining
- Edge runtime v1.0 demo: sub-5ms prediction + tiered inhibition
- Evidence chain format finalised + cryptographic signing
- HCII2026 camera-ready paper submitted (deadline 20 Mar 2026)
- 2–3 deployed field pilots with measurable safety outcomes
- Portable .avm profile marketplace (v0)
- Category-leading latency + reliability benchmarks published
- Non-provisional patent filed (AU + US)