Data & AI Engineering

Start with the hands-on work of SQL, design pipelines that stay safe when they fail, and build the internal parts of an LLM application yourself. By the end, when someone reports that the data looks wrong, you can narrow down with numbers which stage went astray.

코스

• SQL in Practice — From SELECT to the execution plan, all run by hand
• Data Pipelines — Fail, rerun, and the result must be the same
• Apache Spark — The answer to a slow job is in the plan and the event log — Measure shuffles, joins and skew in local-mode Spark with the event log
• Apache Hadoop — Stand up and run HDFS and YARN in one pod — Blocks, permissions, quotas and MapReduce by hand on pseudo-distributed HDFS and YARN
• Apache Flink — Running Streams on a Real Engine — Verify watermarks, state and checkpoints from the engine's own output
• ClickHouse — A Columnar Analytics Database from the Inside — Read sort keys, parts and merges through the numbers in system tables
• LLM Engineering — From the tokeniser to RAG evaluation, from the ground up
• AI Agents — A Graph, Not a Model — Stop conditions, failure paths, traces. Where a demo and a product part ways
• Transformers — Compute Attention By Hand — Confirm with numbers why you divide by √d
• The AI Diet Gone Wrong — Build a real INT8 model and measure accuracy and performance
• Last Week Had a Better Model. Nobody Can Find It — Build the run ledger, registry and promotion gate by hand
• The agent dropped my database — Write an MCP server with the standard library and guard it three ways