FDE Mastery Program
Build Forward Deployed Engineers.
The Forward Deployed Engineer is 2026's fastest-growing applied-AI role — job postings are up ~800%. It takes two things in equal measure: strong engineering and working shoulder-to-shoulder with the client. The people who get the offer can judge whether AI's output is actually right, and turn a vague problem into a clear plan. Every track runs one real client project end to end and finishes with a supervised capstone that scores how ready each person is to deploy.
3routes in
110–536hours, by route
13–67days, by route
10engagement milestones
90%hands-on or proctored
These are what one person does on their route — not a sum of all three (nobody takes every route). Every hour is live teaching, a hands-on lab, or a proctored assessment; a training day is 8 hours.
What you get
Job-ready people, not certificates.
In plain terms — here is what this program gives your business.
Fill the role you can’t hire for
The Forward Deployed Engineer is 2026’s hottest AI role — and brutal to hire. Build it from engineers you already have.
Proof, not attendance
Everyone delivers a real client project and passes a supervised capstone. You see what they can actually ship — not a completion certificate.
Ready for the real job
They rehearse the actual hiring loop and finish with one clear readiness score you can trust before you deploy them.
Mapped to your team
Three routes — for new grads, lateral movers, and internal transfers — scoped to where your people already are.
Three doors in
One standard. Three entry points.
Everyone finishes at the same bar: a supervised client project, a review panel, and a readiness score. What changes is where you start — and how much you can skip by proving you already have the skill.
The engagement
MERIDIAN — one real client project, start to finish.
Everyone builds for the same (made-up) client: NovaFreight, a 4,000-person logistics company that wants an AI assistant to handle shipping problems, reroute deliveries and answer customers — built into its existing software, behind the company login. At every step you hand in real work a panel reviews; nobody advances just for showing up. Logistics is just the shared setting — the skills are identical in healthcare, finance or retail, and for a client cohort we rebuild MERIDIAN around your own domain.
1
Discovery & decomposition
Decomposition memo + draft SoW for NovaFreight
Pass-gate — Panel accepts scope; CIRCLES rubric ≥ 70%
A: M9B: B5C: C1
2
Production repo
Full-stack app + tests + CI + operator README
Pass-gate — Code-review rubric pass
A: M1B: B0 (waived by diagnostic)C: C0 (waived by diagnostic)
3
Unified data layer
Pipeline unifying TMS + CRM extracts into a queryable model
Pass-gate — DQ checks green; correctness spot-audit
A: M2B: B0 (waived)C: (waived)
4
Customer-cloud infra
Terraform-deployed stack, least-privilege IAM, CI/CD
Pass-gate — Security checklist pass
A: M3B: B0 (waived)C: C3
5
Production RAG
RAG over the 100k-doc ops corpus at p95 SLO
Pass-gate — Retrieval metrics + latency SLO met
A: M4B: B1C: C2
6
Agent workflow
Multi-agent rerouting/comms workflow, 2 tools, guardrails
Pass-gate — Task-completion + guardrail coverage thresholds
A: M5B: B1C: C2
7
Eval suite & observability
Golden set, LLM-as-Judge, tracing, quality dashboard
Pass-gate — Eval-design review pass (whiteboard + build)
A: M6B: B2C: C2
8
Enterprise integration
SSO + legacy TMS API integration + runbook
Pass-gate — Integration review + incident drill pass
A: M7B: B3C: C3
9
Architecture defense & exec demo
10–15-min defense + C-suite demo to skeptical panel
Pass-gate — Design rubric + panel score ≥ threshold
A: M8/M9B: B4/B5C: C1/C4
10
Capstone handoff
Full engagement package: deployed system, evals, runbook, playbook
Pass-gate — Capstone panel → DRI score issued
A: CAPB: CAPC: CAP
Curriculum
Every session, every lab, every hour.
Expand any module to see its sessions, the submittable lab each one produces, and how the time is delivered.
Where the time goes
LaunchpadGroundBuild the StackApplied AIProve & OperateConsult & LeadProve It
How the time is delivered
LAB · 321 h (60%)PA · 166 h (31%)ILT · 44 h (8%)SP · 5 h (1%)
M0 — Orientation — What an FDE actually is6 h · 3 sessions
A-M0-01
The FDE role DNA
Explain the FDE mandate (Palantir 'Deltas' origin) and how it differs from SWE / SA / Sales Engineer.
Submittable labMap 5 real FDE JDs onto a skills radar.
ILT2 h
A-M0-02
Market map & the FDE ladder
Navigate the 5 employer segments, role aliases, and the FDE-1 → FDE-2 → Lead ladder.
Submittable labBuild a personal target-employer map (10 companies, 3 segments).
SP2 h
A-M0-03
T-shaped baseline & gap plan
Self-assess against the 7-cluster FDE competency model and commit a gap plan.
Submittable labProctored baseline scorecard + written gap plan.
PA2 h
L1 — Computing launchpad — from college to a working engineer's toolkit34 h · 8 sessions
A-L1-01
Linux & the terminal I
Navigate, manage files/permissions, and chain commands with pipes.
Submittable labComplete a 40-task terminal obstacle course.
LAB4 h
A-L1-02
Linux II & shell scripting
Automate routine work with small bash scripts.
Submittable labWrite 5 scripts (log grep, backup, batch rename, cron, health check).
LAB4 h
A-L1-03
How the internet works
Explain HTTP, DNS, TLS and what happens when you hit Enter on a URL.
Submittable labTrace and annotate a full request with curl + devtools.
ILT3 h
A-L1-04
DSA essentials I
Solve array, string and hashmap problems at screening level.
Submittable lab12 timed problems (easy → medium).
LAB5 h
A-L1-05
DSA essentials II
Apply two-pointer, sliding-window, stack/queue patterns.
Submittable lab12 timed pattern problems.
LAB5 h
A-L1-06
DSA essentials III
Traverse trees and graphs (BFS/DFS) at screen level.
Submittable lab10 timed tree/graph problems.
LAB5 h
A-L1-07
Modern dev workflow
Drive an IDE, a debugger and AI-assisted coding responsibly.
Submittable labDebug a broken project end-to-end with breakpoints + AI pair.
LAB4 h
A-L1-08
Launchpad checkpoint
Prove terminal + DSA + workflow basics under time pressure.
Submittable labProctored basics screen; gate to Phase 1.
PA4 h
L2 — Professional foundation — communicate like a professional from day one14 h · 4 sessions
A-L2-01
Written communication
Write crisp emails, status updates and short docs.
Submittable labRewrite 5 bad emails; draft a weekly status update.
LAB3 h
A-L2-02
Spoken communication & GD
Speak in structured points; hold your own in group discussion.
Submittable labTwo GD rounds + a 3-minute structured talk, peer-scored.
LAB3 h
A-L2-03
Aptitude & reasoning gym I
Solve quantitative and logical reasoning at placement-test pace.
Submittable lab40-question timed set with review.
LAB4 h
A-L2-04
Aptitude gym II + checkpoint
Consolidate speed; clear the aptitude gate.
Submittable labProctored aptitude checkpoint.
PA4 h
M1 — Production software engineering — paced from zero60 h · 13 sessions
A-M1-01
Python fundamentals I
Write correct Python: syntax, control flow, functions.
Submittable lab30 graded exercises.
LAB5 h
A-M1-02
Python fundamentals II
Use collections, comprehensions, files and exceptions fluently.
Submittable labBuild a CSV-crunching CLI from scratch.
LAB5 h
A-M1-03
Python OOP & modules
Model problems with classes; organise code into modules.
Submittable labRefactor the CLI into a small OO package.
LAB5 h
A-M1-04
Python to production I
Add typing and packaging — code others can install and read.
Submittable labType-annotate + package the project; pass mypy.
LAB4 h
A-M1-05
Python to production II
Ship with error handling, structured logging, config and CLIs.
Submittable labAdd logging/config/CLI; fail gracefully.
LAB4 h
A-M1-06
Testing & TDD
Drive code with pytest, coverage gates and CI hooks.
Submittable labReach 85% coverage with meaningful tests.
LAB5 h
A-M1-07
Git & collaboration
Work branch-PR-review like a client team expects.
Submittable labFull PR cycle incl. review comments on a partner's code.
LAB4 h
A-M1-08
SQL for customer data I
Join, aggregate and profile unfamiliar schemas.
Submittable labAnswer 10 business questions on a raw ops schema.
LAB5 h
A-M1-09
SQL for customer data II
Use CTEs and window functions confidently.
Submittable lab15 window/CTE problems on the ops schema.
LAB5 h
A-M1-10
SQL screen (timed)
Perform under CodeSignal-style time pressure.
Submittable labProctored timed SQL screen.
PA3 h
A-M1-11
TypeScript full-stack essentials
Build a typed Node API + React front-end.
Submittable labShip a 2-screen React app on a Node/Express API.
LAB6 h
A-M1-12
APIs — build & consume
Design REST endpoints, handle auth basics, consume third-party APIs.
Submittable labWrap a flaky third-party API behind a resilient endpoint.
LAB5 h
A-M1-13
Milestone build — operable app
Integrate M1: app + tests + CI + a README a customer could operate.
Submittable labPP-2: full-stack app graded by code-review rubric.
PA4 h
M2 — Data engineering foundations38 h · 9 sessions
A-M2-01
ETL/ELT & pipeline thinking
Choose batch vs streaming and design an ingestion plan.
Submittable labWhiteboard an ingestion plan for a 3-source landscape.
ILT3 h
A-M2-02
PySpark I — distributed compute
Explain the Spark execution model; first transformations.
Submittable labProcess a 10 GB event log with PySpark.
LAB5 h
A-M2-03
PySpark II — messy data
Clean, dedupe and conform dirty data at scale.
Submittable labUnify inconsistent customer records across two systems.
LAB5 h
A-M2-04
Data modeling & lakehouse
Model warehouse/lakehouse tables AI apps can query.
Submittable labDesign a star schema + Delta tables for an ops domain.
LAB5 h
A-M2-05
Ingesting the real world
Pull from SFTP drops, ERP extracts and rate-limited APIs.
Submittable labBuild a connector for an SFTP CSV drop with schema drift.
LAB4 h
A-M2-06
Orchestration basics
Schedule and monitor pipelines with Airflow/dbt.
Submittable labOrchestrate the M2 pipeline as a DAG with alerts.
LAB4 h
A-M2-07
Data quality & validation
Add checks that catch bad data before the client does.
Submittable labExpectation suites + quarantine flow.
LAB4 h
A-M2-08
Practice clinic — pipeline debugging
Diagnose and fix broken pipelines under guidance.
Submittable labFix 5 sabotaged pipelines against the clock.
LAB4 h
A-M2-09
Milestone — unified client model
Deliver a clean, queryable model from two messy datasets.
Submittable labPP-3: pipeline graded on correctness + DQ checks.
PA4 h
M3 — Cloud & deployment infrastructure38 h · 9 sessions
A-M3-01
Cloud core services
Map compute/storage/network across AWS, GCP and Azure.
Submittable labStand up equivalent stacks in two clouds (free tier).
ILT4 h
A-M3-02
Docker
Containerise services with sane images and compose files.
Submittable labContainerise the M1 app; multi-stage build under 200 MB.
LAB4 h
A-M3-03
Kubernetes (read-level+)
Deploy, inspect, debug and roll back workloads on K8s.
Submittable labDeploy to a cluster; break it; diagnose from logs/events.
LAB5 h
A-M3-04
Terraform & IaC
Express infrastructure as reviewable, repeatable code.
Submittable labTerraform the app's full stack from zero.
LAB5 h
A-M3-05
CI/CD pipelines
Automate build-test-deploy with environment promotion.
Submittable labPipeline: PR → staging → prod with approvals.
LAB4 h
A-M3-06
IAM, VPC & least privilege
Configure identity and network boundaries in a customer account.
Submittable labLock the deployment to least-privilege roles; prove it.
LAB4 h
A-M3-07
Deploying into a customer's cloud
Follow safe-deployment checklists, secrets and access hygiene.
Submittable labMock customer-account deployment with a checklist.
LAB4 h
A-M3-08
Practice clinic — break/fix drills
Recover broken deployments calmly and methodically.
Submittable lab5 timed break/fix scenarios (IAM, DNS, quota, image, secret).
LAB4 h
A-M3-09
Milestone — infra ship
Deploy the M1 app via Terraform + CI/CD with least-privilege IAM.
Submittable labPP-4: deployment review + security checklist.
PA4 h
M4 — Applied LLMs & RAG in production45 h · 10 sessions
A-M4-01
LLM APIs in anger
Use OpenAI/Claude/Gemini APIs: function calling, streaming, long context.
Submittable labBuild a function-calling service across two providers.
LAB5 h
A-M4-02
Prompt engineering & versioning
Write robust prompts and manage them like code.
Submittable labVersion + A/B two prompt families on a fixed task set.
LAB4 h
A-M4-03
Embeddings & vector stores
Choose embedding models; run pgvector/Pinecone/Weaviate.
Submittable labBenchmark 3 embedding models on domain retrieval.
LAB4 h
A-M4-04
RAG I — chunking strategies
Engineer chunking that fits the corpus, not the tutorial.
Submittable labCompare 4 chunking strategies on retrieval quality.
LAB5 h
A-M4-05
RAG II — hybrid search & reranking
Add BM25 hybrid retrieval and rerankers to lift precision.
Submittable labAdd hybrid + reranker; quantify the precision lift.
LAB5 h
A-M4-06
Fine-tune vs prompt
Argue the cost/quality trade-off; pick correctly per case.
Submittable labDecision memo on 3 client scenarios.
ILT3 h
A-M4-07
Cost & latency engineering
Meet token-cost and p95 latency budgets.
Submittable labCut the RAG system's cost 40% without quality loss.
LAB4 h
A-M4-08
RAG debugging clinic
Diagnose bad answers: retrieval vs generation vs data.
Submittable labTriage 10 failing queries to root cause and fix.
LAB4 h
A-M4-09
Guided studio — RAG build week
Assemble the full corpus → index → serve path with a mentor on call.
Submittable labMentor-supported build time on the MERIDIAN corpus.
LAB6 h
A-M4-10
Milestone — production RAG
Ship RAG over a 100k-doc corpus hitting a p95 latency SLO.
Submittable labPP-5: graded on retrieval metrics + latency SLO.
PA5 h
M5 — Agentic systems & tool use38 h · 8 sessions
A-M5-01
Agent patterns
Apply ReAct, planning and tool-use patterns appropriately.
Submittable labImplement the same task as ReAct vs plan-and-execute.
ILT4 h
A-M5-02
Orchestration with LangGraph
Build stateful agent graphs with retries and branches.
Submittable labBuild a 4-node agent graph with checkpointing.
LAB5 h
A-M5-03
MCP servers & agent skills
Expose tools via MCP; package skills agents can load.
Submittable labWrite an MCP server for an internal API + one skill.
LAB5 h
A-M5-04
Multi-agent & sub-agents
Design researcher/actor/editor sub-agent hierarchies.
Submittable labBuild a 3-sub-agent workflow with a supervisor.
LAB5 h
A-M5-05
Guardrails & human-in-the-loop
Constrain agents with approvals, policies and safe tools.
Submittable labAdd HITL approval + policy guardrails.
LAB5 h
A-M5-06
Routing & state
Route by policy; persist agent state across sessions.
Submittable labAdd policy routing + durable state.
LAB4 h
A-M5-07
Guided studio — agent build
Wire the rerouting/comms workflow with a mentor on call.
Submittable labMentor-supported build time on the MERIDIAN agent.
LAB5 h
A-M5-08
Milestone — agent workflow
Ship a multi-agent workflow with 2 external tools + guardrails.
Submittable labPP-6: task-completion rate + guardrail coverage.
PA5 h
M6 — Evaluation & observability — the differentiator36 h · 8 sessions
A-M6-01
Eval-driven delivery
Explain why evals decide FDE offers and gate deployments.
Submittable labWrite an eval plan before touching the system.
ILT4 h
A-M6-02
Regression & eval suites
Build golden sets and regression suites that catch drift.
Submittable lab150-case golden set for the M4 RAG system.
LAB5 h
A-M6-03
LLM-as-Judge
Design, calibrate and de-bias judge prompts.
Submittable labCalibrate a judge against human labels to κ ≥ 0.7.
LAB5 h
A-M6-04
The six metric families
Measure quality, hallucination, safety, latency, cost, governance.
Submittable labInstrument all six on the agent workflow.
LAB5 h
A-M6-05
Observability & tracing
Trace every hop with LangSmith/Braintrust-class tooling.
Submittable labAdd tracing + a quality dashboard to M5.
LAB5 h
A-M6-06
A/B testing & prompt QA
Run controlled prompt/model experiments in production.
Submittable labA/B two model configs with significance testing.
LAB4 h
A-M6-07
Guided studio — eval suite build
Assemble the full suite with a mentor on call.
Submittable labMentor-supported suite build for MERIDIAN.
LAB4 h
A-M6-08
Milestone — eval suite
Deliver suite + tracing + dashboard; defend on a whiteboard.
Submittable labPP-7: whiteboard + build review.
PA4 h
M7 — Enterprise integration & security30 h · 7 sessions
A-M7-01
Integrating legacy stacks
Bridge REST/GraphQL into ERP/CRM-era systems.
Submittable labIntegrate with a mock SAP-style SOAP/REST hybrid.
LAB4 h
A-M7-02
Enterprise auth
Navigate OAuth, SAML, SCIM and SSO without hand-holding.
Submittable labWire the app behind a mock IdP (SAML + OIDC).
LAB5 h
A-M7-03
Compliance fluency
Speak SOC 2, HIPAA, FedRAMP and data residency credibly.
Submittable labComplete a mock security-review questionnaire.
ILT3 h
A-M7-04
Resilience patterns
Apply rate limiting, retry/backoff, circuit breakers, caching.
Submittable labHarden the integration against injected failures.
LAB5 h
A-M7-05
Cross-system incident drill
Debug failures spanning three systems under time pressure.
Submittable labTimed incident: find the fault across app/IdP/API.
LAB4 h
A-M7-06
Guided studio — integration build
Wire SSO + legacy API with a mentor on call.
Submittable labMentor-supported integration on MERIDIAN.
LAB4 h
A-M7-07
Milestone — enterprise ship
Integrate the agent with SSO + a legacy API; write the runbook.
Submittable labPP-8: integration review + incident-response drill.
PA5 h
M8 — System design for AI products26 h · 6 sessions
A-M8-01
LLM system design
Weave token cost and latency budgets into architecture.
Submittable labDesign a support-copilot architecture with explicit budgets.
ILT5 h
A-M8-02
Scaling patterns
Apply caching, batching and queueing to AI workloads.
Submittable labRedesign the RAG system for 50x traffic.
ILT4 h
A-M8-03
Failure modes & mitigations
Enumerate and mitigate AI-system failure modes.
Submittable labFMEA on the agent workflow; top-5 mitigations.
LAB4 h
A-M8-04
Whiteboard drills I
Hold a 10–15-minute architecture defense.
Submittable labMock defense round with rotating challengers.
LAB4 h
A-M8-05
Whiteboard drills II
Defend under pushback: assumptions, alternatives, trade-offs.
Submittable labSecond mock round, harder follow-ups.
LAB4 h
A-M8-06
Milestone — design defense
Defend: shipment-rerouting agent with a 99% eval suite.
Submittable labGraded on the 6-axis design rubric.
PA5 h
M9 — The consulting half — decomposition, discovery & communication37 h · 9 sessions
A-M9-01
Decomposition under ambiguity
Apply the '48-second rule': scope before coding, always.
Submittable labDecompose a one-line brief into a workplan live.
ILT5 h
A-M9-02
Decomposition case drills I
Run classic cases: 911 response, hospital readmissions.
Submittable labTwo timed case drills with rubric scoring.
LAB4 h
A-M9-03
Decomposition case drills II
Run logistics + one cold unseen domain.
Submittable labTwo timed case drills, unseen domains.
LAB4 h
A-M9-04
Discovery & requirements
Turn a vague business need into a concrete technical plan.
Submittable labDiscovery interview with a role-played stakeholder.
LAB4 h
A-M9-05
SoW & project plans
Write scopes of work and plans a client will sign.
Submittable labDraft the SoW for the MERIDIAN engagement.
LAB4 h
A-M9-06
Executive communication
Explain a RAG architecture to a CTO who's never heard of embeddings.
Submittable lab5-minute exec explainer; graded on clarity.
LAB4 h
A-M9-07
Scope, risk & escalation
Negotiate scope and escalate blockers early without drama.
Submittable labScope-creep role-play + escalation memo.
LAB4 h
A-M9-08
Agile delivery & PM tools
Sequence delivery across Jira/Linear/Notion like a client team.
Submittable labPlan a 2-sprint delivery in Linear with risks logged.
SP3 h
A-M9-09
Milestone — client panel
Present the decomposed plan to a skeptical client panel.
Submittable labPP-1/PP-9: CIRCLES rubric + panel score.
PA5 h
M10 — FDE-2 layer — architecture, mentoring & multi-client delivery20 h · 5 sessions
A-M10-01
Architecture ownership & ADRs
Make and document architecture-level calls.
Submittable labWrite 3 ADRs for the MERIDIAN build.
ILT4 h
A-M10-02
Mentoring FDE-1s
Coach code quality and delivery practice through review.
Submittable labReview a junior's PR; deliver structured feedback.
LAB4 h
A-M10-03
Two-client portfolio delivery
Prioritise and context-switch across concurrent engagements.
Submittable labSimulated week across two client backlogs.
LAB4 h
A-M10-04
Accelerators & playbooks
Codify field patterns into reusable assets.
Submittable labTurn the MERIDIAN deployment into a reusable playbook.
LAB4 h
A-M10-05
Milestone — FDE-2 sprint
Mock two-client sprint: arch call + PR review + playbook.
Submittable labArchitecture memo + mentoring simulation, panel graded.
PA4 h
M11 — Interview gym & placement sprint — fresher-specific114 h · 9 sessions
A-M11-01
Resume, portfolio & GitHub
Present the MERIDIAN portfolio so recruiters shortlist it.
Submittable labShip resume + portfolio site + pinned repos, reviewed.
LAB3 h
A-M11-02
Coding gym I (timed screens)
Perform on CodeSignal-style screens at hiring bar.
Submittable labTwo proctored timed screens with review.
PA4 h
A-M11-03
Coding gym II
Close speed/accuracy gaps from gym I.
Submittable labTwo more proctored screens; trend must improve.
PA4 h
A-M11-04
LLM system design mock
Survive the highest-failure interview round.
Submittable labFull mock round with panel feedback.
PA4 h
A-M11-05
Evals deep-dive mock
Whiteboard a judge + regression suite cold.
Submittable labFull mock round with panel feedback.
PA4 h
A-M11-06
Decomposition case mock
Scope an ambiguous enterprise brief live.
Submittable labFull mock round with panel feedback.
PA4 h
A-M11-07
Behavioural & HR prep
Tell ownership stories with structure (STAR) and calm.
Submittable labRecord + review 6 behavioural answers.
LAB3 h
A-M11-08
Full loop simulation
Run the entire FDE hiring loop end-to-end in one day.
Submittable labScreen → coding → design → evals → case → behavioural; composite score.
PA8 h
A-CAP-01
Capstone engagement
Run the full lifecycle solo over 10 dedicated working days: discovery → SoW → build → integrate → evals → deploy → C-suite demo → handoff.
Submittable labPP-10: elected brief delivered end-to-end for a simulated client; judged on adoption, not demo polish.
PA80 h
Where the time goes
Reframe & VerifyAI Stack SprintEnterprise & DesignConsult & Lead
How the time is delivered
LAB · 72 h (48%)PA · 60 h (40%)ILT · 18 h (12%)
B0 — Role reframe & entry diagnostic6 h · 2 sessions
B-B0-01
SWE → FDE mindset shift
Reframe from feature delivery to embedded outcome ownership.
Submittable labRewrite a past project as an FDE engagement narrative.
ILT2 h
B-B0-02
Entry diagnostic
Prove production code, SQL and cloud baseline (waives foundations).
Submittable labProctored diagnostic: code + SQL + deploy task.
PA4 h
B1 — Applied AI stack sprint32 h · 8 sessions
B-B1-01
LLM APIs & prompting
Function calling, streaming, long context; prompts managed like code.
Submittable labTwo-provider function-calling service + versioned prompts.
LAB4 h
B-B1-02
Embeddings & vector stores
Select embeddings; run pgvector/Pinecone in anger.
Submittable labBenchmark 3 embedding models on domain retrieval.
LAB4 h
B-B1-03
RAG build I — chunking
Engineer corpus-fit chunking.
Submittable labCompare 4 chunking strategies on quality.
LAB4 h
B-B1-04
RAG build II — hybrid & rerank
Lift precision with hybrid search + rerankers.
Submittable labQuantify the reranker precision lift.
LAB4 h
B-B1-05
Agents & orchestration
Build stateful agent graphs with tools.
Submittable lab4-node LangGraph agent with retries.
LAB4 h
B-B1-06
MCP & sub-agents
Expose tools via MCP; design sub-agent hierarchies.
Submittable labMCP server + 3-sub-agent workflow.
LAB4 h
B-B1-07
Cost & latency engineering
Hit token-cost and p95 budgets.
Submittable labCut system cost 40% without quality loss.
LAB4 h
B-B1-08
Sprint milestone
Ship RAG + agent system to spec.
Submittable labPP-5/PP-6 combined build, graded.
PA4 h
B2 — Evaluation & observability20 h · 5 sessions
B-B2-01
Eval-driven delivery
Gate everything on evals; plan them first.
Submittable labEval plan for the sprint system.
ILT4 h
B-B2-02
LLM-as-Judge & golden sets
Build calibrated judges and regression suites.
Submittable lab150-case golden set + judge at κ ≥ 0.7.
LAB4 h
B-B2-03
Six metric families
Quality, hallucination, safety, latency, cost, governance.
Submittable labInstrument all six on the sprint system.
LAB4 h
B-B2-04
Tracing & dashboards
Full-hop tracing + quality dashboards.
Submittable labLangSmith tracing + dashboard live.
LAB4 h
B-B2-05
Milestone — eval suite
Deliver the suite; defend it on a whiteboard.
Submittable labPP-7: build + whiteboard review.
PA4 h
B3 — Enterprise integration refresh12 h · 3 sessions
B-B3-01
Enterprise auth & SSO
OAuth, SAML, SCIM, SSO against a mock IdP.
Submittable labWire the system behind Keycloak (SAML + OIDC).
LAB4 h
B-B3-02
Compliance & resilience
SOC 2/HIPAA/FedRAMP fluency; resilience patterns.
Submittable labMock security questionnaire + failure-injection hardening.
LAB4 h
B-B3-03
Milestone — enterprise ship
SSO + legacy-API integration with runbook.
Submittable labPP-8: review + incident drill.
PA4 h
B4 — System design for AI products12 h · 3 sessions
B-B4-01
Design with budgets
Token cost + latency budgets in architecture.
Submittable labDesign a copilot with explicit budgets.
ILT4 h
B-B4-02
Defense drills
10–15-minute architecture defenses.
Submittable labTwo mock rounds, rotating challengers.
LAB4 h
B-B4-03
Milestone — design defense
Defend the rerouting-agent design.
Submittable lab6-axis rubric, panel graded.
PA4 h
B5 — The consulting half24 h · 6 sessions
B-B5-01
Decomposition under ambiguity
Scope before coding — the 48-second rule.
Submittable labLive decomposition of a one-line brief.
ILT4 h
B-B5-02
Case drills
911 / readmissions / logistics cases, timed.
Submittable labThree rubric-scored drills.
LAB4 h
B-B5-03
Discovery & SoW
Vague need → signed scope.
Submittable labDiscovery interview + MERIDIAN SoW.
LAB4 h
B-B5-04
Executive communication
Explain embeddings to a CTO.
Submittable lab5-minute exec explainer, graded.
LAB4 h
B-B5-05
Scope, risk & escalation
Negotiate scope; escalate early.
Submittable labScope-creep role-play + escalation memo.
LAB4 h
B-B5-06
Milestone — client panel
Present the plan to a skeptical panel.
Submittable labPP-1/PP-9: CIRCLES + panel score.
PA4 h
B6 — FDE-2 delivery leadership44 h · 4 sessions
B-B6-01
Architecture ownership & ADRs
Own architecture calls; write AD.
Submittable lab3 ADRs for the sprint system.
ILT4 h
B-B6-02
Mentoring & multi-client
Coach FDE-1s; run two concurrent engagements.
Submittable labPR review + simulated two-client week.
LAB4 h
B-B6-03
Milestone — playbook
Codify the deployment as a reusable playbook.
Submittable labPlaybook + architecture memo, panel graded.
PA4 h
B-CAP-01
Capstone engagement
Full lifecycle on an elected brief, compressed.
Submittable labPP-10: end-to-end delivery for a simulated client → DRI.
PA32 h
Where the time goes
Bridge DiagnosticConsulting FirstApplied AI EssentialsShadow to Own
How the time is delivered
LAB · 60 h (55%)PA · 40 h (36%)ILT · 4 h (4%)MP · 6 h (5%)
C0 — Source-role bridge diagnostic6 h · 2 sessions
C-C0-01
Bridge diagnostic
Map current role against the FDE DNA; produce a personal gap map.
Submittable labProctored diagnostic + gap map against the bridge matrix.
PA4 h
C-C0-02
Gap plan & mentor pairing
Commit a plan; pair with an FDE mentor for the transition.
Submittable labSigned gap plan reviewed by mentor panel.
MP2 h
C1 — The consulting half — first, because it's the biggest gap20 h · 5 sessions
C-C1-01
Decomposition under ambiguity
Scope before coding — the 48-second rule.
Submittable labLive decomposition of a one-line brief.
ILT4 h
C-C1-02
Case drills
Timed classic cases with rubric scoring.
Submittable labThree drills: 911 / readmissions / logistics.
LAB4 h
C-C1-03
Discovery & SoW
Turn a vague need into a signed scope.
Submittable labDiscovery interview + SoW draft.
LAB4 h
C-C1-04
Executive communication
Explain architecture to non-technical execs.
Submittable lab5-minute exec explainer, graded.
LAB4 h
C-C1-05
Milestone — panel drill
Present a decomposed plan to a skeptical panel.
Submittable labCIRCLES rubric + panel score.
PA4 h
C2 — Applied AI essentials (condensed)28 h · 7 sessions
C-C2-01
LLM APIs & prompting
Function calling, streaming; prompts as code.
Submittable labFunction-calling service + versioned prompts.
LAB4 h
C-C2-02
RAG condensed I
Embeddings, vector stores, chunking.
Submittable labBuild RAG over a domain corpus.
LAB4 h
C-C2-03
RAG condensed II
Hybrid search, reranking, debugging.
Submittable labAdd reranker; triage 5 failing queries.
LAB4 h
C-C2-04
Agents & MCP I
Agent graphs, tools, orchestration.
Submittable lab4-node agent with two tools.
LAB4 h
C-C2-05
Agents & MCP II
MCP servers, sub-agents, guardrails, HITL.
Submittable labMCP server + guardrailed sub-agent flow.
LAB4 h
C-C2-06
Evals condensed
Golden sets, LLM-as-Judge, six metric families.
Submittable lab80-case suite + judge on the build.
LAB4 h
C-C2-07
Milestone — AI build
Ship the RAG + agent + eval package.
Submittable labPP-5/6/7 condensed, graded.
PA4 h
C3 — Customer-cloud deployment & integration16 h · 4 sessions
C-C3-01
Customer-account deployment
Deploy safely into a customer cloud: IaC, IAM, secrets.
Submittable labTerraform deploy with least-privilege proof.
LAB4 h
C-C3-02
Enterprise auth
SSO/SAML/OIDC against a mock IdP.
Submittable labWire the build behind Keycloak.
LAB4 h
C-C3-03
Resilience & runbooks
Harden and document for handoff.
Submittable labFailure-injection hardening + runbook.
LAB4 h
C-C3-04
Milestone — enterprise ship
Integrated, hardened, documented deployment.
Submittable labPP-8 condensed: review + drill.
PA4 h
C4 — Shadow-to-own engagement ladder40 h · 5 sessions
C-C4-01
Shadow a live engagement
Observe a real FDE engagement end-to-end; log the playbook.
Submittable labShadow log + pattern write-up.
LAB4 h
C-C4-02
Co-deliver a workstream
Own one workstream under FDE supervision.
Submittable labDeliver the workstream; supervisor sign-off.
LAB4 h
C-C4-03
Own a workstream
Run a workstream solo incl. client comms.
Submittable labSolo delivery graded on client feedback.
LAB4 h
C-C4-04
Delivery retro
Present the transition portfolio to the mentor panel.
Submittable labRetro + readiness recommendation.
MP4 h
C-CAP-01
Mini-capstone
Deliver a scoped engagement (or simulated brief) end-to-end as the responsible FDE.
Submittable labPP-10 (scoped): discovery → deploy → demo → handoff → DRI.
PA24 h
| Mode | What it means |
|---|---|
| ILT | Instructor-led session (live, cohort) |
| LAB | Hands-on lab with submittable artifact |
| SP | Self-paced (curated resources + exercise) |
| MP | Mentor panel (review / pairing / retro) |
| PA | Proctored assessment (BuildReady / ProctorShield) — feeds PRI → DRI |
Proof
How we decide someone is ready.
Every track ends the same way: a supervised capstone judged by a panel against seven published criteria. The scoring is fixed and public, and each person walks away with a single readiness score — the DRI (Deployment Readiness Index).
Working deployment
System runs in the (mock) customer environment and a simulated user adopts it; judged on outcomes, not demo polish.
25%
Decomposition & scope
Quality of the discovery memo, SoW and workplan; assumptions surfaced; the '48-second rule' honoured.
15%
Evaluation rigor
Golden set, LLM-as-Judge calibration, regression coverage; the six metric families instrumented.
20%
Integration & security
SSO/auth done right, least-privilege IAM, resilience patterns, compliance posture articulated.
15%
Observability & cost
Tracing in place; latency and token-cost budgets defined and met.
10%
Executive communication
C-suite demo lands with a non-technical audience; questions handled with composure.
10%
Handoff & playbook
Runbook a customer could operate; reusable playbook codified for the next engagement.
5%
Track A — Full Embedded Engagement
Complete engagement lifecycle solo: discovery → SoW → data + infra → RAG + agent build → eval suite → enterprise integration → deploy → C-suite demo → handoff.
80 h · 10 working days (8 hrs/day), elected brief (any of the 4)Track B — Compressed Engagement
Same lifecycle, compressed: infra/data foundations pre-provisioned; learner owns decomposition, AI build, evals, integration, defense and handoff.
32 h · 4 working days (8 hrs/day) sprint, elected briefTrack C — First Owned Engagement
A real scoped engagement (or simulated brief if none live): learner is the responsible FDE end-to-end, mentor observes but does not intervene.
24 h · 3 working days of effort, spread over 2–3 weeks alongside live workElectable engagement briefs
Candidates elect a brief, so two people from the same cohort do not present the same system.
Logistics rerouting agent
NovaFreight (default): agentic shipment-exception rerouting with 99% delivery-rate eval suite, TMS integration, exec demo.
Ops/agents-heavyHospital readmissions copilot
Clinical-notes RAG + risk-flag agent for a 12-hospital network; HIPAA posture; clinician-facing explainability.
Regulated/RAG-heavyFintech close-cycle automation
Sub-agent workflow (research/finance/editor) automating month-end close tasks against ERP APIs; SOC 2 narrative.
Integration-heavyCitizen-services RAG (public sector)
Multilingual policy RAG for a government portal; data-residency constraints; FedRAMP-style review.
Governance-heavyTrack C
What your existing engineers already bring.
Internal transfers only cover what they are missing — not the whole syllabus. Each starting role already brings different ground.
| Source role | Carries over | Biggest gaps | Fast-track prescription | Est. hrs |
|---|---|---|---|---|
| Backend / full-stack SWE | Production codeAPIstestinggit |
Consulting halfRAG/agents/evalscustomer-cloud deploy |
C1 fullC2 fullC3 full |
~110 |
| Data engineer | PipelinesSQLSparkwarehousing |
Consulting halfagents & evalsapp-layer + auth |
C1 fullC2 (agents/evals focus)C3 |
~100 |
| DevOps / SRE / Platform | CloudIaCK8sCI/CDincident response |
Consulting halfentire AI stackclient comms |
C1 fullC2 fullC3 light |
~100 |
| Solutions architect / sales engineer | Discoverydemosexec commsscoping |
Production build depthRAG/agents/evals hands-ondeploy |
C1 lightC2 fullC3 fullextra M1 labs |
~110 |
| QA / SDET | Testing disciplineautomationquality mindset |
Consulting halfbuild depthAI stackdeploy |
C1 fullC2 fullC3 fullextra M1 labs |
~120 |
| Tech support / implementation | Customer empathytroubleshootingproduct depth |
Production engineeringAI stackarchitecture |
Pre-work: Track A M1then C1–C4 full |
~150 |
Already covered — credited on evidence
The gap we are closing
What we prescribe for it
Outcome
Rehearsed against the real hiring loop.
Every stage of the interview a Forward Deployed Engineer will actually face is prepared by named modules — not by generic interview coaching.
| Interview stage | What it tests | Prepared by |
|---|---|---|
| Recruiter + technical screen | Fit, communication, timed coding (CodeSignal-style) | A: M1, M9B: B0C: C0/C1 |
| Coding rounds | Production-grade full-stack code; SQL; rate-limiter / job-queue problems | A: M1, M7B: B0, B3C: C0, C3 |
| LLM system design | Token cost, latency budgets, RAG/agents in architecture — highest failure round | A: M4, M8B: B1, B4C: C2 |
| Evals deep-dive | Whiteboard an LLM-as-Judge / regression suite — most-weighted disqualifier | A: M6B: B2C: C2 |
| Decomposition / case study | Ambiguous enterprise brief; scope before solving — lowest pass rate | A: M9, M8B: B5C: C1 |
| Behavioural / customer empathy | Ownership, executive presence, scope negotiation, calm under pressure | A: M9, M10B: B5, B6C: C1, C4 |
Track A modules
Track B modules
Track C modules
The FDE ladder
Rung 1
FDE-1
0–3 yrs · delivers workstreams inside one engagement under guidance; owns quality of own code; learns client comms.
Track A graduate · Track C on completionRung 2
FDE-2
3–6 yrs · owns delivery across two concurrent engagements; mentors FDE-1s; makes architecture calls for smaller systems.
Track A (full) · Track B graduateRung 3
Lead FDE
6+ yrs · owns accounts and the engagement portfolio; grows the team; codifies playbooks; feeds product roadmap.
Experience + FDE-2 layer + portfolio
Next step
Bring us your cohort and we will map it to a track.
Tell us who you are training and where they start. We will come back with the track, the hours, the entry diagnostic and a delivery calendar.
Train the trainer
Can you teach this?
We are always looking for practitioners who can teach to band and assess to a rubric. Send us your profile and we will come back to you about the subjects and bands you could take.
Train the trainer
Apply to teach with Learnlytica
Four fields. We read every profile that comes in.