AI Learning Ramp
A three-times-a-week ramp from GenAI query systems into AI infrastructure, agentic systems, evals, safety, and OpenAI or Anthropic style interviews.
Each new course page should fit one focused hour: a short reading set, a system-design frame, and one interview drill tied back to your BigQuery GenAI query-engine background. The first pass prioritizes AI engineering interview fundamentals before agent breadth.
Your read checkboxes are stored in this browser on this device.
Prefill vs decode, KV cache, continuous batching, PagedAttention, and a BigQuery-style TTFT drill.
vLLM, TGI, TensorRT-LLM, and SGLang engine-selection tradeoffs with SLO and capacity planning.
TTFT, ITL, throughput, prompt caching, and admission-control tradeoffs for mixed interactive and agentic workloads.
Chunking, hybrid retrieval, reranking, freshness, and provenance for grounded answers and agent workflows.
Context selection, compaction, prompt caching, state boundaries, and long-context failure isolation for analytics agents.
DAIL-SQL, DIN-SQL, enterprise Bedrock text-to-SQL architecture, constrained decoding, and permission-safe BigQuery agents.
Workflows, agents, routing, orchestrator-worker, evaluator-optimizer, and handoffs for production analytics agents.
Function calling, tool schemas, auth, sandboxing, retries, receipts, and MCP boundaries for governed analytics agents.
State machines, queues, checkpoints, human approval, replay, and cancellation for long-running analytics agents.
Golden sets, LLM judges, human review, regression tests, release gates, and production feedback for AI query systems.
Traces, tool receipts, state diffs, latency metrics, SLOs, and failure replay for production AI query systems.
Fallbacks, backpressure, rate limits, model drift, rollout strategy, and incident loops for production AI query systems.
The automation will add new course pages here as they are published.