Course Board
This board tracks the work required to turn the current systems coursework into an AI Infrastructure Firmware Engineering specialization. Tickets are Markdown files in the repository, so the plan can evolve alongside lessons, labs, diagrams, reviews, and source history.
The board is the working backlog, not a claim that every listed topic has already been studied or published. A separate GitHub Project view can be used for repository-level planning when synchronization is available.
In progress
Define the specialization architecture and outcomes
Turn the broad AI-infrastructure domain into a coherent dependency graph with explicit learner outcomes.
Expand the OpenBMC and UEFI core track
Continue the existing whole-machine course from its first lesson through host management and fleet operations.
Connect NIC firmware and RDMA to AI infrastructure
Continue the active NIC course into RDMA, congestion, isolation, telemetry, and accelerator-oriented data paths.
Ready
Publish the specialization landing page
Create the public entry point, track map, prerequisites, and honest progress view for the specialization.
Teach rack onboarding and continuous host assurance
Explain how an operator establishes and continuously rechecks the identity, composition, firmware, and configuration of vendor-built racks.
Build the platform identity and attestation sequence
Connect roots of trust, secure boot, measured boot, TPM evidence, component identity, and verifier policy without collapsing them together.
Define the specialization lab environment
Select accessible simulations, commodity hardware, traces, and evidence artifacts that make the system models observable.
Define the AI-assisted authoring and expert-review policy
Make the iterative authoring process credible by documenting source, testing, learner-feedback, and expert-review expectations.
Backlog
Build the PCIe and accelerator lifecycle track
Extend PCIe fundamentals into accelerator discovery, initialization, isolation, telemetry, reset, firmware update, and failure recovery.
Build the fleet firmware lifecycle track
Teach inventory, compatibility, staged rollout, observability, rollback, repair, and retirement across heterogeneous components.
Establish the industry veteran interview series
Add durable practitioner interviews that expose design constraints, operational tradeoffs, and career paths the text alone cannot supply.
Design the rack onboarding and assurance capstone
Integrate management, PCIe, networking, security, and fleet operations in a realistic vendor-rack admission and drift incident.
Blocked
Done
Establish the repository-backed course board
Create a durable ticket schema and a public board that evolves with the specialization.
Ticket conventions
- P0 establishes the foundation or unblocks several other tickets; P1–P3 indicate decreasing urgency.
- Effort is relative: S, M, L, or XL. It is not a deadline or hour estimate.
- A ticket moves to done only when its acceptance criteria are satisfied.
- Browser-local lesson annotations remain the fine-grained review workflow. Tickets capture larger, durable units of course work.