Site / lesson index
Lesson Index. 105 lessons.
Search by title, lede, topic, or section heading. Filter by tier. Open any lesson directly.
Foundations 7 lessons
-
What AI actually is
Compare prompts, check claims against supplied facts and repair a useful draft. Leave with a three-sentence proposal and a fact-check note.
- large language models
- pattern-matching
- hallucination
- prompting loop
- context
- verification
- iteration
- training cutoff
-
Five useful prompts and how to check them
Practice summarizing, rewriting, explaining, comparing and drafting using one fictional source packet.
- summarize
- plain English
- brief in five minutes
- decision frame
- first drafts
- prompt templates
- context-task-constraints
-
Prompt like an Operator
Context, Task, Standards, Evidence: write bounded instructions and inspectable checks, then revise against a specific failure.
- Operator framing
- context
- task
- standards
- few-shot examples
- bad vs good prompts
- structure over tricks
-
When AI is wrong
Choose independent checks for source meaning, numbers, dates and unsupported claims.
- hallucination
- verification
- weak categories
- strong categories
- citations
- training cutoff
- pattern recognition
-
AI for your transition
Draft truthful bullets, compare requirements with evidence and rehearse an interview using a fictional profile.
- military-to-civilian translation
- resume bullets
- job description decode
- interview prep
- networking email
- transition
- callbacks
-
AI vocabulary for choosing a workflow
Separate model, product, harness and environment; match text, reasoning, multimodal and retrieval capabilities to the task.
- LLM
- tokens
- embeddings
- RAG
- fine-tuning
- context window
- agents
- model card
-
Where to go from here
Demonstrate a skill, identify a gap and build four weeks of bounded practice with observable criteria.
- next steps
- intermediate tier
- career paths
- paid tools
- habit-building
- the map
Intermediate 10 lessons
-
Python for AI work
Not general CS. The minimum Python you need to call AI APIs, handle data, and ship something. If you've never written Python, this lesson is for you. If you have, skim and move on.
- Python environment
- venv
- pip
- requirements.txt
- dotenv
- API basics
- scripting
-
AI API requests and safe response handling
Translate a bounded request, inspect completion and usage, and repair a guard that mistakes partial text for success. Required practice is a local simulation; provider examples are optional dated references.
- Anthropic API
- OpenAI API
- authentication
- streaming
- error handling
- retries
- messages
-
Data fundamentals
SQL and pandas. Two tools for getting answers out of tabular data. Practice filtering, joining, cleaning and preserving data lineage.
- SQL basics
- pandas
- CSV
- JSON
- parquet
- data cleaning
- tabular data
-
Prompt engineering at scale
Foundations Lesson 03 taught you Operator prompts. This lesson teaches you to systematize them: templates, variables, structured output, chains. Same Operator framing, more reps, scalable code.
- prompt templates
- variables
- structured output
- JSON mode
- few-shot in code
- chains
- Operator framing
-
Retrieval-augmented generation, from scratch
RAG built end-to-end with no framework. Embeddings, chunks, similarity, prompt assembly. Once you've built one by hand, every framework's RAG is obvious.
- RAG
- embeddings
- chunking
- cosine similarity
- prompt assembly
- no-framework build
- retrieval
-
Vector databases and embeddings
When to graduate from numpy in memory to a real vector store. pgvector, Chroma, Pinecone, Weaviate: when to use what, and how to migrate without rebuilding everything.
- pgvector
- Chroma
- Pinecone
- Weaviate
- indexing strategies
- HNSW
- embeddings at scale
- migration
-
Agents and tool use
What agents actually are versus what the hype says. Function calling. Multi-step reasoning. When agents work, when they fail, and how to keep them on a leash.
- agents
- tool use
- function calling
- multi-step reasoning
- failure modes
- guardrails
- agent loops
-
Evaluation and observability
How to know if your AI app is good, not how it feels. How it measures. Eval sets, regression tests, traces. The Operator move: standards that don't drift.
- eval sets
- regression tests
- LLM-as-judge
- traces
- observability
- metrics
- Operator standards
-
Cost, latency, and trade-offs
Money and time. Token economics, prompt caching, model selection. The Operator's job is to understand what every call costs and how long it takes. Know the numbers before the bill arrives.
- token economics
- prompt caching
- streaming vs batching
- model selection
- response caching
- dedup
- cost discipline
-
Capstone: ship something real
The last lesson of Intermediate. Pick a project. Scope it tightly. Build it end to end. Make a reproducible offline demo a reviewer can inspect. The portfolio you defend in interviews is the capstone.
- capstone project
- scope discipline
- shipping
- spec format
- deliverables
- honest scoping
- portfolio piece
Expert 10 lessons
-
ML and deep learning: enough to be dangerous
The minimum ML vocabulary and intuition you need to read papers, talk to ML scientists, and not get rolled in a design review. We skip the CS major. We pick up the concepts that pay back.
- supervised learning
- loss functions
- gradient descent
- backpropagation
- neural architectures
- training vs inference
- deep learning eval
-
LLM internals: how transformers work
What's actually happening inside the model when you send a prompt. Attention, tokens, embeddings, decoding. The architecture that runs most production AI in 2026, explained without hand-waving.
- transformers
- attention
- tokenization
- embeddings
- decoding
- self-attention
- layer structure
-
Fine-tuning, post-training, and alignment
Choose prompting, lookup/retrieval, tools or adaptation from the observed failure mechanism. SFT, DPO and LoRA solve different parts of the problem.
- fine-tuning decision tree
- SFT
- DPO
- LoRA
- post-training
- dataset quality
- fine-tune ops
-
Production deployment patterns and infrastructure
Going from a notebook to a production system that serves real users. Inference serving (vLLM, TGI), containerization, load balancing, autoscaling, GPU economics. The infrastructure patterns that hold up under federal load.
- vLLM
- TGI
- containers
- Kubernetes
- GPU economics
- autoscaling
- load balancing
- deployment pipeline
-
Monitoring, observability, and incident response
How you know your production AI system is healthy. SLOs and SLIs for AI. Traces, metrics, evals in production. Drift detection. The incident response playbook for AI systems. The discipline that separates engineering from theater.
- SLOs
- SLIs
- logging
- traces
- metrics
- drift detection
- incident response
- postmortems
-
AI security: prompt injection, model risk, supply chain
How attackers will try to break your AI system. Direct and indirect prompt injection. Jailbreaks. Model extraction. Supply chain. Adversarial inputs. The defensive moves that work in 2026, and the ones that don't.
- prompt injection
- indirect injection
- jailbreaks
- model extraction
- data exfiltration
- supply chain
- OWASP LLM Top 10
-
Federal AI compliance: NIST AI RMF from the engineer's seat
Apply GOVERN, MAP, MEASURE and MANAGE to engineering evidence, risk decisions and ownership. Distinguish voluntary framework guidance from actual organizational and authorization requirements.
- NIST AI RMF
- GOVERN MAP MEASURE MANAGE
- SYSTEM_CARD.md
- bias evaluation
- cybersecurity
- continuous compliance
- engineering artifacts
-
ATO and FedRAMP for AI systems
Distinguish cloud-service evidence from agency system authorization, define the exact boundary, and prepare a reviewable synthetic decision record.
- ATO
- FedRAMP
- Impact Levels
- IL2-IL6
- SCAP
- STIGs
- cATO
- OSCAL
-
Procurement from the engineer's chair
Reading SOOs and PWSs like an engineer. Writing the technical volume. Estimating level of effort honestly. Capability statements that get callbacks. The procurement skills senior engineers actually need on the federal side.
- SOO
- PWS
- technical volume
- level of effort
- capability statement
- contracting officer
- color of money
-
Capstone: production AI in a federal context
Nine lessons in. The capstone is the system you can hand to a hiring panel, an evaluator, or a contracting officer and have it defend itself. Architecture, code, evals, security, compliance package, briefing. The portfolio piece that closes the deal.
- federal capstone
- architecture decision records
- evals package
- security review
- compliance package
- briefing
- honest scoping
Annex A: Cleared 7 lessons
-
What cleared AI work actually means
Separate mission, access and technical work. Learn with synthetic examples, bounded practice and evidence-based review.
- cleared AI market
- classification ladder
- DoD buyers
- intel community
- prime contractors
- competition
- starting position
-
OPSEC for AI tools
Decide before disclosure. Learn with synthetic examples, bounded practice and evidence-based review.
- OPSEC
- data spillage
- commercial AI tools
- decision tree
- personal devices
- clearance risk
- historical incidents
-
On-prem and air-gapped AI
Map the whole deployment. Learn with synthetic examples, bounded practice and evidence-based review.
- on-prem
- air-gapped
- open-weights models
- Llama
- Mistral
- GPU hardware
- local tooling
- deployment trade-offs
-
The cleared AI role landscape
Read the work behind the title. Learn with synthetic examples, bounded practice and evidence-based review.
- GS civilian
- prime contractors
- FFRDC
- small business
- compensation bands
- role archetypes
- first interview
-
The certs that count
Choose credentials from requirements. Learn with synthetic examples, bounded practice and evidence-based review.
- DoD 8140
- Security+
- CISSP
- cloud certs
- AI/ML certs
- cert sequencing
- employer-paid training
-
Talking about your work without violating
Make claims both truthful and releasable. Learn with synthetic examples, bounded practice and evidence-based review.
- resume language
- LinkedIn rules
- interview answers
- type of problem framing
- common slips
- non-disclosure
- self-review
-
Where to go from here
Plan learning and authorized onboarding. Learn with synthetic examples, bounded practice and evidence-based review.
- career archetypes
- first 90 days
- compounding moves
- commercial offer trade-off
- 12-month plan
- long view
Annex B: Leadership 7 lessons
-
What you actually need to know about AI as a leader
Describe a capability as a system. Learn with synthetic examples, bounded practice and evidence-based review.
- leader's model of AI
- vendor questions
- model cards
- calibrated skepticism
- translation layer
- judgment
- decision framing
-
Vendor evaluation that doesn't get gamed
Run a reproducible mission comparison. Learn with synthetic examples, bounded practice and evidence-based review.
- RFP
- bake-off
- capability statement
- reference checks
- vendor theater
- evaluation boards
- buyer leverage
-
Writing AI requirements into a contract
Turn outcomes into testable requirements. Learn with synthetic examples, bounded practice and evidence-based review.
- PWS
- SOO
- evaluation criteria
- NIST AI RMF clauses
- ATO timelines
- vendor language
- contract levers
-
Risk management at the leader's altitude
Connect risk to controls and authority. Learn with synthetic examples, bounded practice and evidence-based review.
- NIST AI RMF
- OMB memos
- executive orders
- safety-impacting
- rights-impacting
- five-minute brief
- escalation policy
-
The budget cycle and color of money
Estimate lifecycle work without authorizing spending. Learn with synthetic examples, bounded practice and evidence-based review.
- PPBE
- POM
- color of money
- RDT&E
- OCO
- sustainment math
- continuing resolution
- funding profile
-
Building and managing an AI team
Build coverage around responsibilities. Learn with synthetic examples, bounded practice and evidence-based review.
- org design
- AI roles
- hiring filters
- cleared talent
- retention conversation
- culture
- commercial competition
-
The first 180 days as an AI leader
Sequence evidence before scale. Learn with synthetic examples, bounded practice and evidence-based review.
- listening tour
- quick wins
- strategy memo
- execution rhythm
- political navigation
- 180-day plan
- long arc
Optimize Your Workspace 1 lessons
-
Set up a workspace your AI can navigate
Organize sources, drafts and reviewed notes while checking context, permissions and data paths with synthetic records.
- project folder structure
- root instructions file
- task routing
- context window
- AI memory
- reading and writing memory
- one fact per file
- agent builder prompt
Academy: Orientation & Placement 3 lessons
-
Mission, roles, and the learning contract
Set honest expectations, select a defensible pathway, and establish the learning/data contract.
- orientation & placement
- mission
- roles
- learning
- contract
-
Data handling and tool authorization gate
Make the stop/proceed decision before a prompt, upload, log, or action.
- orientation & placement
- data
- handling
- tool
- authorization
- gate
-
Diagnostic and safe workspace setup
Create an isolated least-privilege workspace and use diagnostics without waiving safety.
- orientation & placement
- diagnostic
- safe
- workspace
- setup
Academy: AI-Aware Operator 8 lessons
-
What modern AI systems do
Build an accurate system-level mental model of capability and limitation.
- ai-aware operator
- what
- modern
- systems
-
Safe use before prompting
Turn authorization and minimization into a daily operating habit.
- ai-aware operator
- safe
- before
- prompting
-
Prompt like an operator: Context, Task, Standards, Evidence
Write bounded, testable instructions without treating prompts as security controls.
- ai-aware operator
- prompt
- like
- operator
- context
- task
- standards
- evidence
-
Verification, sources, and calibrated confidence
Practice claim-level verification for consequential output.
- ai-aware operator
- verification
- sources
- calibrated
- confidence
-
Research, summarization, and drafting
Produce source-bounded work with visible provenance.
- ai-aware operator
- research
- summarization
- drafting
-
Decision support and human oversight
Expand analysis without delegating accountability.
- ai-aware operator
- decision
- support
- human
- oversight
-
Tools, multimodal inputs, and agent boundaries
Recognize how retrieval and action expand risk and design least-privilege boundaries.
- ai-aware operator
- tools
- multimodal
- inputs
- agent
- boundaries
-
Foundations field exercise
Demonstrate safe end-to-end work on an unfamiliar bounded task.
- ai-aware operator
- foundations
- field
- exercise
Academy: Applied AI Practitioner 8 lessons
-
Workflow audit and use-case selection
Select bounded work where AI can create measurable value without exceeding risk.
- applied ai practitioner
- workflow
- audit
- use-case
- selection
-
Reusable CTSE workflows and quality standards
Convert a successful prompt into a controlled reproducible package.
- applied ai practitioner
- reusable
- ctse
- workflows
- quality
- standards
-
Data preparation, provenance, and records
Prepare information while preserving ownership, lineage, minimization, and lifecycle.
- applied ai practitioner
- data
- preparation
- provenance
- records
-
Source-grounded research and retrieval
Preserve claim-level traceability while resisting stale, missing, contradictory, or malicious content.
- applied ai practitioner
- source-grounded
- research
- retrieval
-
Documents, spreadsheets, and multimodal work
Produce accessible office artifacts with medium-specific verification.
- applied ai practitioner
- documents
- spreadsheets
- multimodal
- work
-
Workflow evaluation and operational metrics
Determine whether a workflow improves work using representative evidence and predeclared gates.
- applied ai practitioner
- workflow
- evaluation
- operational
- metrics
-
Team rollout, change control, and incident handling
Move a workflow to controlled team use with ownership, rollback, and evidence preservation.
- applied ai practitioner
- team
- rollout
- change
- control
- incident
- handling
-
Applied workflow capstone
Integrate authorization, design, evidence, operations, handoff, and defense.
- applied ai practitioner
- applied
- workflow
- capstone
Academy: AI Application Builder 12 lessons
-
Command line, Git, Python, and environments
Create reproducible work from a clean directory.
- ai application builder
- command
- line
- python
- environments
-
HTTP, JSON, APIs, secrets, and failure handling
Interpret request/response behavior and status codes.
- ai application builder
- http
- json
- apis
- secrets
- failure
- handling
-
Provider-neutral generation APIs
Define a minimal provider-neutral contract.
- ai application builder
- provider-neutral
- generation
- apis
-
Data engineering for AI applications
Validate mixed-format inputs against explicit schemas.
- ai application builder
- data
- engineering
- applications
-
Software quality, testing, and safe telemetry
Choose unit, integration, and contract tests.
- ai application builder
- software
- quality
- testing
- safe
- telemetry
-
Structured outputs and bounded tools
Validate structured outputs independently of the model.
- ai application builder
- structured
- outputs
- bounded
- tools
-
Retrieval baseline from first principles
Chunk content with stable source identifiers.
- ai application builder
- retrieval
- baseline
- first
- principles
-
Production RAG: authorization, hybrid search, and grounding
Filter by authorization before content reaches generation.
- ai application builder
- production
- authorization
- hybrid
- search
- grounding
-
Evaluation design and release gates
Design representative and adversarial cases.
- ai application builder
- evaluation
- design
- release
- gates
-
Agents and tool use under control
Implement an explicit decide-act-observe state loop.
- ai application builder
- agents
- tool
- under
- control
-
Deployment, operations, and cost evidence
Package configuration separately from code.
- ai application builder
- deployment
- operations
- cost
- evidence
-
Builder capstone and technical defense
Integrate a reproducible bounded application.
- ai application builder
- builder
- capstone
- technical
- defense
Academy: Production AI Engineer 12 lessons
-
Machine learning foundations for system builders
Distinguish learning regimes and their evidence needs.
- production ai engineer
- machine
- learning
- foundations
- system
- builders
-
Transformer and inference internals
Trace tokenization, attention, residual flow, and decoding.
- production ai engineer
- transformer
- inference
- internals
-
Adaptation, fine-tuning, and post-training decisions
Choose among prompting, retrieval, tools, tuning, and replacement.
- production ai engineer
- adaptation
- fine-tuning
- post-training
- decisions
-
Production architecture and inference serving
Translate workload and SLOs into capacity.
- production ai engineer
- production
- architecture
- inference
- serving
-
AI threat modeling and adversarial testing
Map assets, actors, trust boundaries, and abuse paths.
- production ai engineer
- threat
- modeling
- adversarial
- testing
-
Identity, authorization, privacy, and data protection
Implement least-privilege authorization for users and services.
- production ai engineer
- identity
- authorization
- privacy
- data
- protection
-
Evaluation engineering and statistical evidence
Design stratified comparisons tied to risks and populations.
- production ai engineer
- evaluation
- engineering
- statistical
- evidence
-
Observability, safe logging, and incident response
Define signals across application, model, retrieval, and tools.
- production ai engineer
- observability
- safe
- logging
- incident
- response
-
Reliable workflows and agents
Model durable state explicitly.
- production ai engineer
- reliable
- workflows
- agents
-
Performance, reliability, and resource governance
Test latency, throughput, availability, and degradation.
- production ai engineer
- performance
- reliability
- resource
- governance
-
Secure supply chain and continuous delivery
Inventory code, model, data, dependency, and image artifacts.
- production ai engineer
- secure
- supply
- chain
- continuous
- delivery
-
Production capstone, incident exercise, and board
Operate a production-like synthetic service against SLOs and gates.
- production ai engineer
- production
- capstone
- incident
- exercise
- board
Academy: Federal AI Mission Practitioner 9 lessons
-
Mission, information, and authorization boundaries
Information and authority matrix. Apply the method to synthetic evidence, explain the result and review its limitations.
- federal ai mission practitioner
- mission
- information
- authorization
- boundaries
-
OPSEC and approved AI use
Disclosure and aggregation. Apply the method to synthetic evidence, explain the result and review its limitations.
- federal ai mission practitioner
- opsec
- approved
-
FedRAMP 20x, cloud impact levels, and authorization boundaries
Offering-specific boundary review. Apply the method to synthetic evidence, explain the result and review its limitations.
- federal ai mission practitioner
- fedramp
- cloud
- impact
- levels
- authorization
- boundaries
-
RMF and the Authority to Operate
RMF implementation and assessment. Apply the method to synthetic evidence, explain the result and review its limitations.
- federal ai mission practitioner
- authority
- operate
-
Continuous authorization and DevSecOps evidence
Continuous attributable evidence. Apply the method to synthetic evidence, explain the result and review its limitations.
- federal ai mission practitioner
- continuous
- authorization
- devsecops
- evidence
-
Federal AI governance and high-impact use
Analyze the function before choosing a policy label.. Apply the method to synthetic evidence, explain the result and review its limitations.
- federal ai mission practitioner
- federal
- governance
- high-impact
-
On-premises, disconnected, and open-weights operations
Disconnected configuration operations. Apply the method to synthetic evidence, explain the result and review its limitations.
- federal ai mission practitioner
- on-premises
- disconnected
- open-weights
- operations
-
Federal acquisition artifacts for engineers
Acquisition comments engineers can defend. Apply the method to synthetic evidence, explain the result and review its limitations.
- federal ai mission practitioner
- federal
- acquisition
- artifacts
- engineers
-
Federal specialization capstone and board
Traceable capstone defense. Apply the method to synthetic evidence, explain the result and review its limitations.
- federal ai mission practitioner
- federal
- specialization
- capstone
- board
Academy: AI Program Leader 7 lessons
-
Leader's model of AI capability and accountability
System accountability. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai program leader
- leaders
- model
- capability
- accountability
-
Use-case portfolio, value, and stop decisions
Value and counterfactuals. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai program leader
- use-case
- portfolio
- value
- stop
- decisions
-
Vendor evaluation and mission bake-off
Reproducible mission comparison. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai program leader
- vendor
- evaluation
- mission
- bake-off
-
Requirements, test strategy, and acceptance
Executable acceptance requirements. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai program leader
- requirements
- test
- strategy
- acceptance
-
AI risk governance and incident leadership
Incident facts and recovery gates. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai program leader
- risk
- governance
- incident
- leadership
-
Funding, lifecycle resources, and fiscal role discipline
Lifecycle quantity planning. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai program leader
- funding
- lifecycle
- resources
- fiscal
- role
- discipline
-
Team, operating model, and 180-day leadership practicum
Operating coverage and evidence gates. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai program leader
- team
- operating
- model
- 180-day
- leadership
- practicum
Academy: AI Career Evidence 4 lessons
-
Role families and evidence-based gap analysis
Requirement-to-evidence mapping. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai career evidence
- role
- families
- evidence-based
- analysis
-
Resume, profile, and interview translation with OPSEC
Truthful contribution and disclosure. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai career evidence
- resume
- profile
- interview
- translation
- opsec
-
Portfolio, technical narrative, and credential selection
Reproducible portfolio packaging. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai career evidence
- portfolio
- technical
- narrative
- credential
- selection
-
Career campaign and evidence defense
A claim must fit the artifact that supports it.. Apply the method to synthetic evidence, explain the result and review its limitations.
- ai career evidence
- career
- campaign
- evidence
- defense
No lessons match your search. Try a different word, or turn a tier filter back on.