Five phases, end to end.

BY THE END

1. 01Write production-quality Python that passes mypy --strict and has ≥80% pytest coverage by default — not as a stretch goal but as a professional baseline.
2. 02Read an unfamiliar TypeScript or Node.js codebase and write a clean, typed integration adapter within a single focused work session.
3. 03Recognize and apply the ten algorithmic patterns that recur in FDE interviews and production data work, with the discipline to handle edge cases and explain the approach to a non-technical stakeholder.
4. 04Use Git's interactive rebase, bisect, and stash effectively; read and diagnose GitHub Actions CI failures; and write code reviews that are direct, specific, and actionable.
5. 05Build a mental model of an unfamiliar codebase in under four hours using a repeatable reading strategy, and produce a technical brief that stands alone without the source.

MODULES

1. M-0101

   Python Production Fluency

   Type-annotated, tested, idiomatic Python at the level a Palantir or Anthropic FDE is expected to write on day one in a client environment.

   CLASS III
2. M-0102

   TypeScript for Polyglot FDEs

   Functional TypeScript competence — enough to read client codebases, write typed integration adapters, and not break type safety under deadline pressure.

   CLASS II
3. M-0103

   Data Structures and Algorithms at FDE Depth

   The ten algorithmic patterns that recur in FDE interviews and production data work — calibrated to the actual difficulty of the Palantir coding round, not LeetCode Hard.

   CLASS II
4. M-0104

   Dev Workflow: Git, CI, and Code Review as a Skill

   Git past add-commit-push, GitHub Actions fundamentals, and code review as a communication discipline — the workflow fluency FDEs need to operate in foreign repositories under deadline pressure.

   CLASS II
5. M-0105

   Reading and Debugging at Speed

   Build a mental model of an unfamiliar codebase in under four hours. Debug a production failure in a client environment with stakeholders watching. Both are trainable skills with repeatable methods.

   CLASS II

BY THE END

1. 01Write window functions, CTEs, and recursive queries, and read a query plan well enough to explain why a query is slow and how to fix it.
2. 02Choose between row stores and columnar engines (DuckDB, ClickHouse, BigQuery/Snowflake) based on workload, and run analytical queries against a columnar store on a real dataset.
3. 03Reason about replication, consistency models, and partitioning at DDIA depth — enough to diagnose stale reads, split-brain, and hot partitions on a client call.
4. 04Operate the AWS primitives an FDE actually touches (IAM, S3, Lambda, RDS, VPC networking) with least-privilege defaults.
5. 05Containerize a workload and deploy it to Kubernetes with health checks, resource limits, and a sane Dockerfile.
6. 06Provision infrastructure declaratively with Terraform and instrument a service with OpenTelemetry traces, metrics, and logs.
7. 07Manage secrets and enforce least privilege so that a deployment leaks nothing — no hardcoded credentials, no over-scoped roles.

MODULES

1. M-0201

   SQL Past the Basics and the Columnar Shift

   Window functions, CTEs, and query plans, plus when to reach for a columnar engine instead of a row store — the data layer every deployment starts from.

   CLASS III
2. M-0202

   Distributed Systems Literacy at DDIA Depth

   Replication, consistency models, and partitioning deep enough to diagnose stale reads, split-brain, and hot partitions on a live client call.

   CLASS III
3. M-0203

   The AWS Primitives an FDE Actually Touches

   IAM, S3, Lambda, RDS, and VPC networking — the cloud surface you land on at most clients, operated with least-privilege defaults.

   CLASS II
4. M-0204

   Containers and Orchestration

   Docker images built right and Kubernetes deployments with health checks and resource limits — how an FDE ships a workload that survives a client's cluster.

   CLASS III
5. M-0205

   Infrastructure as Code and Observability

   Provision infrastructure declaratively with Terraform and make systems legible with OpenTelemetry logs, metrics, and traces.

   CLASS III
6. M-0206

   Security and Secrets Management

   The discipline that runs through every deployment: managing secrets, enforcing least privilege, and threat-modeling so an FDE engagement leaks nothing.

   CLASS II

BY THE END

1. 01Design production-grade prompts and agent loops using system/user separation, structured outputs, and tool use rather than ad-hoc single-shot prompting.
2. 02Build a grounded retrieval pipeline with deliberate chunking, hybrid search, and reranking, and justify each design choice against the customer's data.
3. 03Treat evaluation as a first-class engineering discipline: author task-specific eval suites, wire them into CI, and use eval deltas to drive product decisions.
4. 04Apply explicit decision criteria for prompting vs. retrieval vs. fine-tuning, and defend the tradeoff to a non-technical stakeholder.
5. 05Operate an LLM system in production: version prompts and datasets, instrument observability for non-deterministic behavior, and catch regressions before customers do.

MODULES

1. M-0301

   Prompt Engineering Past the Surface

   System vs. user prompts, structured outputs, tool use, and the anatomy of an agent loop — the prompting skills that survive contact with production.

   CLASS II
2. M-0302

   Retrieval Design for Grounded Systems

   Chunking, hybrid search, reranking, and contextual retrieval treated as a data-systems problem — the layer that grounds LLM answers in the customer's data.

   CLASS III
3. M-0303

   Evaluation as a First-Class Discipline

   The load-bearing module. Author task-specific eval suites with Inspect AI, OpenAI Evals, Braintrust, and Anthropic's tooling, and treat them as version-controlled code that gates CI.

   CLASS III
4. M-0304

   Fine-Tuning vs. Prompting: The Decision

   Explicit decision criteria for the optimization ladder — prompting, retrieval, then fine-tuning — and how to defend the tradeoff to a customer who wants to fine-tune everything.

   CLASS II
5. M-0305

   MLOps for LLM Systems

   Versioning prompts and datasets, observability for non-deterministic behavior, and regression testing in CI — the operational discipline that keeps a deployment alive after launch.

   CLASS II

BY THE END

1. 01Run structured discovery interviews that distinguish stated requirements from revealed behavior, and translate findings into a feasible/valuable/usable scope.
2. 02Write and speak in Pyramid Principle / SCQA structure so that a busy executive grasps the answer, the supporting logic, and the ask within the first thirty seconds.
3. 03Build a RACI and influence map for a deployment that identifies the champion, the economic decision-maker, and the likely blocker before scoping work.
4. 04Diagnose and write a recovery plan for a deployment where the technology works but adoption has stalled for organizational rather than engineering reasons.
5. 05Operate a disciplined field-to-core feedback loop, converting field observations into product- and model-roadmap input the core org can act on.

MODULES

1. M-0401

   Customer Discovery as an Engineering Discipline

   Treat requirements-gathering like systems work: structured interviews, hypotheses, and a hard separation between what clients say and what they actually do.

   CLASS III
2. M-0402

   Pyramid Principle, MECE, and SCQA

   Barbara Minto's frameworks for answer-first writing and speaking, so an executive gets the conclusion and the ask before they lose patience.

   CLASS III
3. M-0403

   Stakeholder Mapping and Executive Communication

   RACI, the champion/decision-maker/blocker triad, and writing status updates an executive can act on without a meeting.

   CLASS II
4. M-0404

   The Political Layer of a Stalled Deployment

   Diagnosing and recovering deployments where the technology works but adoption doesn't — the most common FDE failure mode and the least documented.

   CLASS III
5. M-0405

   Field-to-Core Feedback Loops

   What an FDE owes the product and model org: turning field reality into roadmap input the core team can actually act on.

   CLASS II

BY THE END

1. 01Assemble a portfolio of applied-AI work that demonstrates measurable workflow impact and eval-driven iteration, not feature demos.
2. 02Write a resume and LinkedIn profile calibrated to FDE-specific vocabulary and the conventions for describing ambiguous, client-facing work.
3. 03Distinguish and prepare for all four FDE interview formats — coding, system design, customer scenario, and behavioral — and articulate how each differs from a SWE loop.
4. 04Evaluate and negotiate an FDE compensation package across base, equity, on-call, and deployment-travel components.
5. 05Construct a 30/60/90 plan grounded in the FDE deployment lifecycle, anchored to concrete first-quarter deliverables.

MODULES

1. M-0501

   Portfolio Construction for FDE Hiring

   Frame your capstone and earlier builds as evidence of deployed value — measurable workflow impact, eval-driven iteration, and ambiguity navigated — not as feature demos.

   CLASS II
2. M-0502

   Resume and LinkedIn Calibrated for FDE Recruiters

   Translate your experience into FDE-specific vocabulary, describe ambiguous client work without overclaiming, and pass both the keyword screen and the human read.

   CLASS II
3. M-0503

   Networking and Outreach That Doesn't Read Like a Template

   Identify the people who can route you past the resume pile and reach them with specific, evidence-led messages that earn a reply.

   CLASS II
4. M-0504

   The Four FDE Interview Formats

   Prepare for coding, system design, customer scenario, and behavioral rounds — and learn exactly how each diverges from a standard SWE loop.

   CLASS III
5. M-0505

   Salary Negotiation for FDE Comp Structures

   Evaluate and negotiate an FDE package across base, equity, on-call, and deployment-travel components, which behave differently from a pure-SWE offer.

   CLASS II
6. M-0506

   The FDE 30/60/90 Plan

   A first-quarter plan synthesized from the FDE deployment lifecycle, with each window anchored to a concrete, demonstrable artifact.

   CLASS III