Automated Incident Diagnosis
A default incident-investigation path that turns a PagerDuty signal into selected operational reads, streamed progress, and a final response an engineer can act on.
A default incident-investigation path that turns a PagerDuty signal into selected operational reads, streamed progress, and a final response an engineer can act on.
A fleet health experience that turns existing telemetry into repository-health scores, then opens deeper investigation only when an engineer chooses a repository or question.
An evaluation workflow that uses expert references to score eligible production answers while keeping offline regression checks and release judgment separate.
A compatible TaskSets compiler path that lets a team express real parallelism inside an existing pipeline without changing serial workflows that do not opt in.
Operational work across a shared event-data path, from Kafka and Spark enrichment through batch dataset production and consumer-specific delivery.
A diagnostic practice for locating a streaming pressure source before changing partitioning, consumer parallelism, or intake rate.
A recovery workflow that identifies an affected hour, waits for complete input, limits revision starts, and verifies the repaired slice against its source.
A scheduled logistics-risk integration that prepares Kafka events in Spark, applies configured MLflow inference, and delivers the result to operational APIs and dashboards.
A repeatable Azure Databricks recovery foundation, built with Terraform and exercised through a failover drill before workloads are expected to move.
Agents can accelerate investigation without becoming the source of truth.
Agent research should expose what was found, what failed, and what remains unanswered.
Model creativity can sit inside a runtime whose task, context, tools, and side effects remain explicit.
Local SQLite storage with hybrid retrieval, retention rules, and a consolidation loop.
A two-tool local MCP workflow publishes session recaps and module pages with a lightweight graph.
A zero-dependency scanner with focused supply-chain and artifact checks.
A local pipeline records crawl, translation, narration cleanup, generation, and packaging artifacts by phase.
Curriculum, source, card-generation, and release checks are organized as a cross-agent skill.