The themes that shape ASTR-1
Each theme corresponds to an active research direction at Neognos Tech and a part of the ASTR-1 architecture.
Battery-first materials discovery
Scope and priorities are anchored to battery materials, not generic chemistry. Pipelines, gates, and benchmarks are tuned to the failure modes that matter for cells.
Quantum-inspired generative algorithms
QIGA explores candidate compositions and structures using quantum-inspired formulations. The algorithms run on classical hardware; no quantum hardware is required.
Hybrid property prediction
Combining learned models with physics-informed estimators to predict target properties at multiple fidelity levels while making cost and confidence explicit.
Uncertainty-aware validation
Decisions are gated on calibrated uncertainty. High-impact or low-confidence candidates are escalated to higher-fidelity validation rather than accepted by default.
Dopant and composition repair
When a candidate is promising but flawed, ASTR-1 proposes targeted dopants and composition adjustments aimed at the specific weakness rather than discarding the candidate.
Failure database and learning flywheel
Negative results — predictions that did not hold up, syntheses that failed, candidates that degraded — are captured and reused to improve future routing decisions.
Synthesis planning
Candidate materials that survive the workflow are paired with candidate synthesis routes and precursors, considering practical lab constraints and feasibility.
Permission-gated physical testing
Physical and robotic experimentation is opt-in and gated, not assumed. ASTR-1 hands off to the lab only when the evidence and the operator both justify it.
Decision-first evaluation
We evaluate ASTR-1 on the quality of its decisions — advance, repair, reject, validate, or plan for synthesis — not solely on candidate throughput.