Tag

MLOps

Articles tagged “MLOps” on Ryshe.

Aerospace LLM Evaluation with Domain Expert Scoring

Combine automated checks with expert scoring, traceable tests, and release gates to evaluate aerospace LLMs and prevent unsafe deployments.

Oct 29, 2026 7 min read

Secure Azure OpenAI traffic with HTTPS on every hop: enforce TLS, configure private endpoints and DNS, and test allowed/blocked paths.

Oct 29, 2026 8 min read

Measure MRR@10 for Azure AI Search: define relevance, build judged queries, compare retrieval setups, and track production metrics.

Oct 29, 2026 8 min read

Compare BLEU with learned metrics (COMET, BLEURT, BERTScore) for translation QA; combine scores with deterministic checks and human review.

Oct 29, 2026 8 min read

Don't treat connection metadata as proof; verify identity, network path, retrieval, rotation and audits before production.

Oct 29, 2026 6 min read

Use source-aware COMET scoring as a QA gate to auto-approve, flag, and route translations with auditable thresholds.

Oct 29, 2026 14 min read

Treat EASA alignment as an early task: define AI role and boundaries, map requirements, use runtime guards, and retain traceable evidence.

Oct 29, 2026 18 min read

How AI scorecards spot supplier risk - focusing on data quality, explainable models, and governance for reliable action.

Oct 29, 2026 12 min read

AI methods to detect and stop free-trial abuse: anomaly detection, identity clustering, behavior scoring, velocity checks, pattern matching.

Oct 29, 2026 13 min read

Connect telemetry, maintenance, quality, and supplier data in Azure to score risk, trigger alerts, and power role-based reviews.

Oct 29, 2026 17 min read

Use Azure to unify plant data, model takt and bottlenecks, run what-if scenarios, and route only approved line changes to operators.

Oct 29, 2026 15 min read

Treat Azure AI workflow evaluation as continuous: batch tests, shadow runs, human review, threshold tuning, and cross-system validation.

Oct 29, 2026 11 min read