Technology (Including AI)
Technology (Including AI)
- Problem and requirements clarity: What is being built, for whom, for what constraints, and what “success” means.
- System boundaries & interfaces: What’s inside/outside; dependencies; APIs; data flows.
- Architecture soundness: Modularity, coupling, fault isolation, scalability assumptions.
- Correctness & validity: Does it do the intended thing under stated conditions? (For AI: validity of task framing and ground truth assumptions.)
- Reliability & resilience: Failure tolerance, redundancy, disaster recovery, graceful degradation.
- Performance & cost: Latency, throughput, compute, storage, unit economics, cost predictability.
- Security posture: Threat model, attack surface, access control, secrets, supply-chain risk.
- Privacy & data governance: Data minimization, retention, consent, compliance, leakage risk.
- Human factors & usability: Operator burden, workflow fit, alert fatigue, interpretability needs.
- Operational readiness: Observability, logging, monitoring, on-call playbooks, incident response.
- Change management: Versioning, backward compatibility, rollout/rollback, migrations.
- Validation & testing strategy: Unit/integration/e2e; load testing; chaos testing; red teaming.
- AI-specific model behavior: Generalization, robustness, calibration, hallucination/error modes, distribution shift.
- AI evaluation integrity: Dataset quality, leakage, benchmark gaming, reproducibility, adversarial testing.
- Safety & misuse risk: Dual-use, escalation pathways, harmful outputs, abuse vectors.
- Governance & accountability: Who owns outcomes, failure liability, auditability, decision rights.
- Vendor/third-party dependency risk: Lock-in, SLA fragility, roadmap risk.
- Maintenance & lifecycle: Technical debt, model drift, retraining cadence, deprecation strategy.