Engineering / advanced
Operate AI at scale
Design and operate data, training, inference, reliability, and security systems for AI.
Your route through engineering
Follow the path or begin with a question. Every lesson is open.
Your place, at your pace. Progress is optional and saved only in this browser when available.
Read / explore / reason
Inside this module
Module 01 advanced
Operate AI at scale
Design and operate data, training, inference, reliability, and security systems for AI.
- 01
Data pipelines and lineage
Build reproducible training datasets by tracking identities, event time, transformations, and publication boundaries.
26 min - 02
Distributed training parallelism
Choose data, sharded, tensor, and pipeline parallelism by reasoning about memory, communication, and synchronization.
31 min - 03
Inference serving and batching
Reason about prefill, decoding, KV-cache memory, admission control, and latency when serving language models.
29 min - 04
Accelerators and scheduling
Place accelerator workloads using multidimensional resource requests, topology, fairness, and interruption costs.
27 min - 05
Checkpoints and storage
Design training recovery around complete state, crash-consistent publication, validated restore, and checkpoint overhead.
28 min - 06
Reliable messaging and backpressure
Design replay-safe consumers, bounded queues, and explicit overload behavior for asynchronous AI workflows.
29 min - 07
Replication and consistency
Distinguish replica availability, linearizability, transaction isolation, and durable consensus when coordinating infrastructure.
32 min - 08
SLOs and incident response
Define user-centered reliability, calculate error-budget burn, and respond to incidents with evidence and reversible mitigation.
29 min - 09
Security and isolation
Constrain untrusted models, data, and tools with explicit trust boundaries, least privilege, provenance, and layered isolation.
30 min - 10
Capacity planning and cost
Combine workload distributions, latency-tested capacity, failure headroom, and explicit unit economics into a defensible scaling plan.
31 min