Product Roadmap
From Steel Reasoning Engine
to Deeper Plant Autonomy
Three phases. Twenty-four months. The trajectory from a hardened steel reasoning engine and active steel-plant engagements to a productized steel platform and deeper autonomy across the steelmaking shop.
Timeline
Three phases of the steel roadmap
Phase 1
Foundation
Months 1–6
Harden the steel reasoning engine across crane, furnace, caster and ladle, and advance the active steel-plant engagements.
Phase 2
Productization
Months 7–12
Move from engagement to a repeatable steel product — operator tooling, alerting, plant integration, and additional steel sites and lines.
Phase 3
Deeper Plant Autonomy
Months 13–24
Broaden the same architecture across more of the steelmaking shop, with tighter endpoint and quality reasoning and self-improving steel physics models.
Phase 1 · Months 1–6
Foundation
Harden the steel reasoning engine across crane, furnace, caster and ladle, and advance the active steel-plant engagements.
Reasoning Engine
- Harden the Monitoring → Prediction → Reasoning → Decision cycle across crane, furnace, caster and ladle
- Strengthen confidence scoring and uncertainty handling on each reasoning agent
- Make recognition-based clearance and endpoint reasoning robust to noisy plant signals
- Keep deterministic geometric barriers in charge — the engine directs attention, it never owns the stop
Steel Physics
- Deepen BOF kinetics — two-zone converter modeling for endpoint reasoning
- Heat-transfer and thermal models for furnace and ladle states
- Caster solidification reduced-order model for shell and quality reasoning
- Clearance and crane kinematics for overhead-bay geometry
Steel-Plant Engagements
- HeatVue — EAF operating-pulpit video wall, advancing the Rev F commercial offer
- LVS — ladle visualization for the DE bay (ID, lining condition, lift/landing safety)
- HookVision — recognition-based crane anti-collision fusing camera and radar
- BOF & caster surrogates — research-stage reduced-order and ML models for endpoint and quality reasoning
Phase 2 · Months 7–12
Productization
Move from engagement to a repeatable steel product — operator tooling, alerting, plant integration, and additional steel sites and lines.
Operator Tooling
- Repeatable operator and pulpit dashboards built from the active engagements
- Visibility walls for furnace shell, lance-interaction and ladle-handling zones
- Routed recommendations that surface to the right operator at the right moment
- Co-design with pulpit operators so the tooling reflects how the shop actually runs
Alerting & Integration
- Clearance alerting for overhead-crane bays as a packaged capability
- Endpoint and quality reasoning alerts for BOF and caster operations
- Integration with plant PLC / SCADA for telemetry and signal exchange
- Operator feedback loops that refine reasoning thresholds over time
Additional Sites & Lines
- Generalize the architecture to additional cranes, furnaces and casters within steel
- Package the steel physics layer for reuse across new lines
- Extend to further steel sites beyond the initial engagements
- Steel systems reasoning conversations with additional operators
Phase 3 · Months 13–24
Deeper Plant Autonomy
Broaden the same architecture across more of the steelmaking shop, with tighter endpoint and quality reasoning and self-improving steel physics models.
Shop-Wide Coverage
- Broaden the same reasoning architecture across more of the steelmaking shop
- Connect crane, furnace, caster and ladle reasoning into one coherent shop view
- Extend recognition-based clearance management across additional overhead bays
- Keep humans and deterministic barriers in the loop for every safety-critical action
Endpoint & Quality
- Tighter BOF and caster endpoint reasoning for steel quality outcomes
- Closer coupling between thermal, solidification and metallurgical endpoint models
- Earlier detection of deviation from a heat's or a bay's normal state
- Quality reasoning that operators can inspect and trust before acting
Self-Improving Physics
- Steel physics models that refine as observed prediction accuracy is measured
- BOF kinetics, caster ROM and heat-transfer surrogates improving from plant data
- Reasoning transparency so model updates remain auditable
- A steadily sharper steel-specific physics layer underpinning every agent
How We Iterate
Roadmap principles
Earn the engagement first
Each phase has to prove itself inside a real steelmaking shop before the next is built. The active engagement is the qualification, not the pitch deck.
Steel physics constrains the scope
We extend across crane, furnace, caster and ladle only when the steel physics is reliable enough to ground the reasoning. No speculative expansion beyond steel.
Operator feedback is the compass
Pulpit and crane operators surface what the shop actually needs — not what product managers assumed. The roadmap updates accordingly.
Building on a steelmaking floor?
Co-creation engagements across crane, furnace, caster and ladle — open now