FDE training: from engineer to forward-deployed engineer
A forward-deployed engineer takes an AI-assisted solution from a client conversation to production, and proves it works. We teach that on your own repositories, in eight weeks, with something shipped every week.
The curriculum
- AI-native workflow: Working with coding agents in a codebase you did not write. Ships: A feature in an unfamiliar repo, reviewed.
- Agents in delivery and quality: Agents that write and maintain tests and run in CI. Ships: A regression suite in CI, flake rate measured.
- LLM fundamentals: Tokens, context, structured output, tool calling, cost, and the least-machinery ladder. Ships: A tool-calling service, cost per call tracked.
- Evals: Golden sets, code checks, calibrated judges and a gate that fails the build. Ships: A harness that blocks a bad PR.
- Agents and RAG in production: Measured retrieval, agent memory, routing, tool design, guardrails and red-teaming. Ships: An agent with a red-team suite.
- Production and root cause: Tracing, cost and latency budgets, and root-cause analysis. Ships: An RCA on a seeded failure.
- The forward-deployed part: Discovery, scoping, demoing, and saying so when the problem does not need AI. Ships: A mock client discovery and a proposal.
- Capstone: One problem taken from conversation to a working system with evidence. Ships: Demo, eval results, handover document.
How we prove it works
- Before and after on the same real tasks: cycle time, flake rate, eval pass rate and cost.
- A capstone with eval evidence and a handover document.
- A 30, 60 and 90-day scorecard.
Plan a cohort