AI-SDLC Course Series Series Overview
Start Here 30-40 min No Prerequisites

AI-SDLC Overview — See the Whole Picture Before You Start

Before you open the first course, understand what you're building, how the pieces fit together, and where AI fits into every phase of software development. This is your map for the entire series.

Learning Objectives

By the end of this overview, you will be able to:

  • Explain in one sentence what the AI-SDLC series builds and why it matters
  • Name the 6 learning tracks and the specific AI tools each track covers
  • Describe the UCC Lien Risk platform — the unified domain project every course uses
  • Map AI tools (Gemini, Claude Code, OpenSpec, MCP, agents) to the SDLC phase where each one fits
  • Choose your personal starting point based on your role and experience level

What Is an AI-Assisted SDLC? (And Why Should You Care?)

The Software Development Lifecycle (SDLC)The structured process every engineering team follows: gather requirements, design the system, write code, test it, deploy it, and maintain it. AI-SDLC means using AI tools to accelerate every one of those phases — not just code generation. has existed since the 1960s. The phases haven't changed: plan, design, build, test, deploy, monitor. What has changed is that AI can now assist — or even lead — in every single one of those phases.

Most developers think "AI for coding" means a tab-completion tool that finishes your lines faster. That's a tiny slice. The real transformation is when AI is embedded into every handoff point of your development process: writing the requirements ticket, designing the API contract, generating the data pipeline, writing the test suite, reviewing the pull request, and auditing the deployed system for bias. That's what this series teaches.

The Construction Analogy

Traditional SDLC: A construction crew that manually drafts blueprints, hand-measures each beam, calls suppliers to check inventory, files permits on paper, and inspects work with clipboards. Each handoff takes days. Errors are caught late, when they're expensive.

AI-SDLC: The same crew, but now an AI assistant drafts the blueprints from a one-paragraph spec, checks building codes automatically, suggests structural improvements before a single nail is hammered, flags code that violates safety standards the moment it's written, and generates the inspection report from sensor readings — all in hours, not days.

The key insight: The humans still make all the decisions. The AI handles the mechanical transformation — from idea to spec, from spec to code, from code to tests, from tests to report. You define what to build; AI handles the how faster than you could alone.

Technical Definition

An AI-assisted SDLC combines three capabilities across every development phase: (1) IDE intelligence — AI that reads your codebase and suggests contextually-aware completions (Gemini Code Assist, GitHub Copilot); (2) Agentic workflows — AI that can take multi-step actions like reading specs, writing code, running tests, and iterating (Claude Code, AI Agents); and (3) Contract-driven design — AI that generates and validates API contracts before implementation begins (OpenSpec).

Crucially, none of these tools require you to abandon your existing workflow. They fit into the SDLC phases you already know — they just compress the time and improve the quality at each phase.

Traditional SDLC vs AI-SDLC — Side by Side

The animation below shows the same task — building a new API endpoint — done two ways. On the left, the traditional approach: meetings, manual spec writing, slow handoffs. On the right, the AI-assisted approach: from requirement to tested code in a fraction of the time. Watch the comparison play out step by step.

Animation: Traditional vs AI-Assisted Development

Without AI (Traditional)

📋 Sprint planning meeting (2 hrs)
⚠ Requirements unclear — back to product team
✍ Manually write API spec (1 day)
⚠ Spec drift — frontend built against old version
🔧 Implement endpoint (3 days)
⚠ Test coverage: 40%, prod bug found in week 2
~1.5 weeks · Many handoffs

With AI-SDLC

🤖 Jira AI drafts ticket from 2-sentence brief
📄 OpenSpec generates API contract from ticket
⚡ Gemini implements endpoint against contract
✅ Testcontainers validates against real DB
🔍 Claude Code reviews, catches edge cases
✅ Deployed — 95% test coverage, zero drift
~2 days · AI at every step
Why It Matters in Production

At a commercial bank processing 50,000 UCC filings per month, a 3-day implementation cycle means a new compliance feature takes 6 weeks from idea to deployment (planning, design, build, test, review, deploy). With an AI-assisted pipeline, the same feature ships in under a week. At scale, that velocity advantage compounds: your team ships 8-10x more features per quarter — or the same features at a fraction of the headcount.

Common Misconceptions About AI-SDLC

"AI will replace developers." — Incorrect. AI eliminates the mechanical transformation work (boilerplate, test scaffolding, spec writing). What remains is judgment: architecture decisions, product tradeoffs, debugging complex emergent behavior. Those tasks get more important, not less.

"This is just fancier autocomplete." — Far beyond that. Autocomplete suggests the next token. AI-SDLC agents read your entire codebase, understand the data model, generate full features aligned to your patterns, write the tests, and flag their own errors. The jump from autocomplete to agentic coding is like the jump from a calculator to a spreadsheet.

"You need to change your whole workflow." — No. Every tool in this series is designed to slot into git-based development, standard CI/CD, and existing Jira/Confluence workflows. The changes are additive, not replacements.

"AI-generated code is unsafe or unreliable." — Only if you have no review process. This series dedicates an entire track to Responsible AI: bias detection, audit trails, output validation, and governance. AI-generated code under review is safer than human code with no tests.

The Domain: UCC Lien Risk Intelligence

Every course in this series uses the same real-world platform as its subject: a UCC Lien Risk Intelligence System for commercial credit decisions. This isn't a toy to-do app. It's the kind of platform that a commercial bank's underwriting team or a factoring company actually runs.

Here's what it does: When a business applies for a commercial loan, lenders need to know what liens already exist against the borrower's assets. A UCC filing (Uniform Commercial Code Article 9)A legal document filed with a US state government when a lender takes a security interest in a borrower's personal property (inventory, equipment, receivables). It's public record, searchable by anyone. Lenders check these before extending credit to understand existing collateral obligations. is a public record that says "this creditor has a claim on these assets." Aggregating, deduplicating, and scoring these filings across all 50 states — in near real time — is the core problem this platform solves.

Why This Domain?

UCC filings are perfect for teaching AI-SDLC because they're messy, unstructured, and inconsistent across states — exactly the kind of real-world data problem that AI tools are designed to handle. "Acme Corporation", "ACME CORP", and "Acme Corp." are three different strings that represent the same legal entity. A deterministic script needs you to enumerate every variant. An AI system reasons about it at runtime. That single problem — entity resolution at scale — threads through every course in the series and gives you a concrete benchmark to measure whether your AI implementation actually works.

The Platform You Will Build

Across all six tracks, you assemble one complete system. Each course adds a new layer. By the end of the series, you have a running, AI-instrumented platform from data ingestion to risk dashboard. Watch the pipeline come together:

Animation: The Platform You Build
📄 Raw Filings 50 states · CSV/XML/JSON
🫨 Bronze Layer PySpark · Parquet · GCS
🫶 Silver Layer Dedup · Standardize · BigQuery
🏅 Gold Layer Risk Scores · Entity Profiles
Spring Boot API REST · OAuth2 · Redis Cache
⚛️ React Dashboard TanStack Query · TypeScript
LayerWhat It ContainsCourse
Raw50,000+ UCC-1, UCC-3, UCC-5 filings — inconsistent formats, duplicates, name variantsDomain Reference
BronzeRaw data preserved exactly as received, partitioned Parquet files on GCS (or local)Data Engineering
SilverDeduplicated (50K → 41K), name-standardized, schema-validated, BigQuery tablesData Engineering
Gold12 entity risk profiles: Acme Corporation risk_score=42 (HIGH), aggregated liensData Engineering
APISpring Boot REST: GET /filings/search, GET /entities/{id}/risk, OAuth2, RedisSpring Boot API
UIFiling search + risk dashboard, TanStack Query, MSW testing, Playwright E2EReact / Angular Frontend
AgentsMCP server exposing the API, AI agent answering "What's Acme Corp's total lien exposure?"MCP + AI Agent

The 6 Learning Tracks

The 22 courses are organized into six tracks. Each track builds on the previous ones, but you can enter at your level — a senior data engineer can skip Foundation and start at Build Track. Watch the tracks light up to see how they connect:

Animation: The 6 Tracks
📚
1. Foundation Domain Reference · Prompt Engineering
⚙️
2. Build Track Data Eng · API · Frontend (GCP + Local)
3. AI Tooling Gemini Code Assist · OpenSpec
🤖
4. Agentic AI MCP Servers · AI Agent
🛠️
5. Project Mgmt Atlassian AI (Jira · Confluence · Rovo)
🛡️
6. Governance Responsible AI · Measuring Adoption

Track Detail

Track 1: Foundation — No prerequisites. The two courses here are not optional prerequisites — they are the multipliers that make every other course 3x faster. The Domain Reference course teaches you UCC filings, lien types, state variations, and the business logic behind risk scoring. Without it, the data models in the Build Track are opaque. The Prompt Engineering course teaches the RCTF framework, context engineering with CLAUDE.md / GEMINI.md, and how to write prompts that generate production-quality code rather than tutorial snippets.

Track 2: Build Track — Full Stack — This is the technical heart of the series. You build every layer of the UCC platform: the Bronze→Silver→Gold data pipeline in PySpark (Data Engineering), the REST API in Spring Boot 3 with PostgreSQL, Redis caching, and OAuth2 (Spring Boot API), and the filing search and risk dashboard UI in React with TypeScript, TanStack Query, and MSW (Frontend). Every course has a GCP (cloud) and a Local (Docker/DuckDB) edition — zero cloud account required to follow along.

Track 3: AI Tooling — These courses teach AI tools that slot into the IDE workflow. Gemini Code Assist covers in-line completions, chat panel queries, Agent Mode for multi-file refactoring, code review assistance, and CI-level metrics collection. OpenSpec covers the spec-first API design workflow: generate an OpenAPI 3.1 contract from requirements, validate implementations against it, and run Pact contract tests to prevent drift. Both courses have Python/TypeScript and Java editions.

Track 4: Agentic AI — The frontier. MCP Servers teaches the Model Context Protocol: expose your UCC Risk API as a set of AI-callable tools (searchFilings, calculateRiskScore, getEntityProfile), connect to Claude Desktop, and deploy to Cloud Run. The AI Agent course teaches you to build a UCC analysis agent from scratch using the Anthropic API tool_use loop, multi-tool orchestration, memory patterns, and multi-agent coordination.

Track 5: Project Management — Atlassian AI is the glue that connects engineering to product management. Jira MCP integration lets AI agents create and update tickets from code changes. Confluence AI generates and updates technical documentation. Rovo agents automate sprint planning, retrospectives, and stakeholder reporting. This course shows how AI-SDLC changes not just the code, but the entire delivery process.

Track 6: Governance — Two courses that are often skipped and should never be. Responsible AI covers bias detection in risk scoring, explainability requirements for adverse action decisions, audit trail design, and the EU AI Act applicability to commercial credit systems. Measuring AI Adoption covers the metrics and dashboards you need to demonstrate ROI to leadership: acceptance rates, time-saved calculations, and quality indicators.

AI at Every SDLC Phase

Here's the insight that ties the whole series together: every AI tool this series teaches maps to a specific phase of the software development lifecycle. It's not a replacement for the process — it's a layer on top of every step. Watch the phases play out and see where each tool enters:

Animation: Where AI Enters the SDLC
1
Requirements & Planning Jira AI drafts tickets from natural language · Rovo suggests acceptance criteria · CLAUDE.md / GEMINI.md defines project context for AI sessions
2
API & System Design OpenSpec generates OpenAPI 3.1 contracts from requirements · Gemini suggests schema improvements · Contract-first prevents consumer/producer drift
3
Implementation Gemini Code Assist provides in-IDE completions aligned to your codebase · Agent Mode refactors multi-file features in one command · Claude Code reads, writes, and verifies code in a loop
4
Testing AI generates Testcontainers integration tests against real PostgreSQL · Pact contract tests validate API consumer-producer contracts · Playwright E2E tests generated from user stories
5
CI/CD & Deployment AI-generated GitHub Actions pipelines · Harness CD pipeline with AI rollback detection · Claude Code generates Dockerfiles and Cloud Run configs from project analysis
6
Governance & Monitoring Responsible AI bias audits on risk scores · Audit trail for adverse action decisions · Adoption metrics dashboard showing AI tool ROI across the team
The Flywheel Effect

Each AI tool in this series generates artifacts that the next AI tool consumes. Jira AI writes the ticket → OpenSpec generates the API contract from the ticket → Gemini implements against the contract → AI generates tests against the implementation → Claude Code reviews and patches → the MCP agent exposes the endpoint to downstream AI systems. This isn't a collection of disconnected tools — it's a pipeline where AI output becomes AI input at each stage. The whole series is designed to teach you how to assemble that flywheel.

Course Roadmap — All 22 Courses at a Glance

Here's every course in the series, organized by track. The number of courses per track reflects the complexity of that concern — Build Track is largest because there are three tech stacks (data, API, frontend) each with cloud and local editions.

Animation: Full Course Roadmap
📚 Foundation
Public Records Domain Prompt Engineering Prompt Eng (Java)
⚙️ Build: Data
Data Eng (GCP) Data Eng (Local) Data Eng OpenSpec (Local)
⚙️ Build: API
Spring Boot API (GCP) Spring Boot API (Local)
⚙️ Build: Frontend
UCC React Frontend UCC Frontend (Local) Angular Frontend Angular Frontend (Local)
✦ AI Tooling
Gemini Code Assist Gemini (Java) OpenSpec OpenSpec (Java)
🤖 Agentic AI
MCP Servers AI Agent
🛠️ Project Mgmt
Atlassian AI
🛡️ Governance
Responsible AI Measuring Adoption
Why Cloud + Local Editions?

Every Build Track course has two editions: a GCP (cloud) edition using Dataproc, BigQuery, Cloud Run, and Cloud SQL — and a Local edition using Docker Compose, DuckDB, and local PySpark. If you have a GCP account, use the cloud edition for production-realistic experience. If you don't (or if your org has restrictions), the Local edition runs identically on Windows 11 + Docker Desktop with zero cloud cost. The skills transfer 100% — only the infrastructure names differ.

Your Learning Path — Pick Your Starting Point

Not everyone starts at the same place. Here are the recommended paths by role. All paths eventually converge on the full platform — the difference is where you enter and which tracks you prioritize:

🌱 New to AI Tools

You write code but haven't used AI tooling in your workflow yet. Start at the beginning and go in order.

Domain Reference → Prompt Eng → Gemini → OpenSpec → Build Track

🔧 Data Engineer

You work in data pipelines but want AI-assisted tooling and agentic data workflows.

Domain Reference → Data Eng → Gemini (data) → MCP → AI Agent

☕ Java / Backend Dev

Spring Boot is your world. You want AI pair programming and contract-driven API development.

Domain Reference → Prompt Eng (Java) → Gemini (Java) → OpenSpec (Java) → Spring Boot API

⚛️ Frontend Dev

You build UIs and want AI-assisted component generation, test writing, and E2E coverage.

Domain Reference → Prompt Engineering → Frontend (React or Angular) → Gemini

🤖 AI / Agentic Focus

You're specifically interested in MCP, agents, and AI tool integration — not the full-stack build.

Domain Reference → Prompt Engineering → MCP Servers → AI Agent → Responsible AI

👔 Tech Lead / Manager

You want to understand AI-SDLC to lead adoption, measure ROI, and make tool decisions.

Domain Reference → Prompt Eng → Atlassian AI → Measuring Adoption → Responsible AI
The Universal First Two Courses

Regardless of your path: start with the Domain Reference course (US Public Records — Commercial Credit Data) and then Prompt Engineering. Every technical course assumes you understand what a UCC-1 filing is, what deduplication means in the context of debtor names, and how to write prompts that produce consistent, production-quality output. These two courses are the multipliers that make everything else faster. They're both beginner-level and have no code prerequisites.

How Courses Connect — The Unified Platform Narrative

This isn't a collection of independent tutorials. Every course adds a real, running component to the same platform. Here's how the handoffs work in practice:

// The course dependency graph (simplified)

Domain Reference"what is a UCC-1 filing, risk_score field, entity dedup logic"
Prompt Engineering"RCTF framework, CLAUDE.md context, chain-of-thought patterns"

Data Engineering uses domain knowledge to:
  ingest_bronze(raw_csv) → deduplicate() → score_risk() → gold_profiles

Spring Boot API exposes gold layer:
  GET /entities/{id}/risk → reads gold_profiles → caches in Redis

React Frontend consumes the API:
  FilingSearch → TanStack Query → RiskDashboard

MCP Server wraps the API as AI tools:
  searchFilings(query) · calculateRiskScore(entityId)

AI Agent orchestrates the MCP tools:
  "What's Acme Corp's total exposure?" → while(needs_tool) → answer

Responsible AI audits the agent:
  bias_check(risk_scores) · audit_trail(adverse_actions)

The data that flows through the Data Engineering course's Bronze→Silver→Gold pipeline is the same data the Spring Boot API queries. The API the Spring Boot course builds is the same API the MCP Server course wraps as AI tools. The MCP server tools are the same ones the AI Agent course orchestrates. Nothing is thrown away between courses — every component is reused and extended.

What Happens If You Skip Courses

Skip Domain Reference: You can still follow the code, but the data model will feel arbitrary. You won't understand why there are three name variants for Acme Corp, why UCC-3 amendments matter, or why the Gold layer is different from the Silver layer. Every course will take 30% longer to follow.

Skip Data Engineering and jump to API: The API course provides a mock dataset, so you can follow along — but you'll be building an API against a data source you don't understand. The queries will seem arbitrary. Take at least the first 2 phases of Data Engineering before starting the API.

Skip straight to AI Agent: The agent course uses the MCP server to call the API. Without building the API, you're connecting to a black box. The agent exercises will work, but you won't understand what you're orchestrating — which means you won't be able to debug it when it fails.

Knowledge Check

Test your understanding of the AI-SDLC series before you dive in. Select the best answer for each question.

Q1: What is the fundamental difference between "AI as autocomplete" and AI-SDLC?

AI-SDLC uses a more powerful model with higher token limits
AI-SDLC integrates AI assistance at every phase of development (planning, design, test, deploy, governance), not just code generation
AI-SDLC replaces human developers entirely with autonomous agents
AI-SDLC is only applicable to Python and JavaScript projects
Correct! AI-SDLC is about embedding AI assistance at every handoff point: requirements → design → implementation → testing → deployment → governance. Autocomplete is just one tool in one phase. The series teaches the full stack of AI tooling across the entire lifecycle.
Not quite. The key distinction is coverage: AI-SDLC means AI at every phase (planning, design, build, test, deploy, governance) — not just code generation. It's about transforming the entire development process, not a single step.

Q2: The UCC Risk platform's data goes through Bronze, Silver, and Gold layers. What is the Silver layer's primary responsibility?

Storing raw filings exactly as received from each state — no transformation
Computing entity risk scores and aggregating lien totals per company
Deduplication, name standardization, and schema validation — turning messy raw data into clean, consistent records
Exposing data via REST API endpoints to the frontend
Correct! Silver is the "clean" layer. Raw Bronze data (50,000+ filings with duplicates, inconsistent names, and format variations) gets deduplicated (50K → 41K), name-standardized ("ACME CORP" → "Acme Corporation"), and schema-validated. Gold then builds on this clean foundation for aggregation and risk scoring.
Not quite. Bronze = raw data preserved exactly. Silver = deduplicated + name-standardized + schema-validated clean data. Gold = aggregated risk profiles and entity scores. API = exposing Gold data to consumers.

Q3: Why are there both GCP and Local editions of every Build Track course?

So learners without a cloud account can still complete every lab using Docker Compose, DuckDB, and local PySpark with zero cloud cost — same skills, different infrastructure names
The GCP and Local editions teach different skills — one is for cloud engineers and one is for on-premises developers
Local editions are for beginners and GCP editions are for advanced learners
The course content is different — GCP covers more advanced topics than the Local edition
Correct! Both editions teach identical skills — the difference is infrastructure only. GCP uses Dataproc, BigQuery, Cloud Run. Local uses DuckDB, Docker Compose, local PySpark. Skills transfer 100%. This ensures no learner is blocked by cloud account requirements or corporate access restrictions.
Not quite. The GCP and Local editions are parallel tracks with identical course content — only the infrastructure differs. The Local edition uses DuckDB + Docker instead of BigQuery + GCP, but teaches the same patterns with the same data.

Q4: Which track is responsible for turning the Spring Boot API into a set of AI-callable tools that an agent can orchestrate?

Track 2 (Build Track) — the API itself is the integration point
Track 3 (AI Tooling) — OpenSpec generates the tool schemas
Track 5 (Project Management) — Atlassian AI connects the API to Jira
Track 4 (Agentic AI) — MCP Servers wraps the API as searchFilings, calculateRiskScore, and getEntityProfile tools
Correct! Track 4's MCP Servers course teaches you to expose the Spring Boot API as Model Context Protocol tools: searchFilings(), calculateRiskScore(), getEntityProfile(). These tools are then what the AI Agent course orchestrates. Without MCP, the agent has no hands — it can reason about UCC risk but can't actually query the data.
Not quite. Track 4 (Agentic AI) is responsible for this. The MCP Servers course teaches how to wrap the Spring Boot API as Model Context Protocol tools that any AI model (Claude Desktop, AI agents) can call. The AI Agent course then orchestrates those tools in a multi-step reasoning loop.

Q5: Why does the series recommend starting with the Domain Reference course, regardless of your role?

It teaches the CSS and HTML design system used in all courses
Every technical course uses UCC filing data models, dedup logic, and business rules — understanding the domain makes the code models obvious rather than arbitrary
It's the only course without code prerequisites and serves as a technical introduction to Python
It contains the Prompt Engineering framework used in all subsequent courses
Correct! Every course in the series uses the same data model: UCC-1 filings, debtor/secured party names, filing dates, collateral descriptions, state variations, and risk scores. Without understanding what a UCC filing IS and why "ACME CORP" vs "Acme Corporation" matters, the deduplication code in Data Engineering and the entity resolution in the AI Agent course will feel meaningless rather than solving a real problem.
Not quite. The domain reference course establishes the business context that all technical courses build on. Every code example — the PySpark dedup job, the Spring Boot entity endpoint, the AI agent query — only makes sense if you understand what UCC filings are and why commercial credit risk scoring depends on them.

Q6: The "flywheel effect" described in this overview refers to which property of AI-SDLC tooling?

AI tools get faster the more you use them due to model fine-tuning on your codebase
Each AI tool reduces errors so that later phases require less testing effort
Each AI tool's output artifact becomes the input for the next AI tool — Jira AI → OpenSpec → Gemini → AI tests → MCP → Agent — creating a compounding chain of AI-generated value
The more AI tools a team adopts, the lower their cloud infrastructure costs become
Correct! The flywheel is about chaining AI-generated artifacts: Jira AI writes the ticket → OpenSpec generates the API contract from the ticket → Gemini implements against the contract → AI generates tests against the implementation → MCP wraps the implementation as agent tools → the AI agent orchestrates it all. Each tool consumes the previous tool's output. That chain is what distinguishes AI-SDLC from "a collection of AI tools."
Not quite. The flywheel refers to AI output becoming AI input across phases: Jira ticket → API contract → implementation → tests → MCP tools → agent orchestration. It's not about model performance or cost reduction — it's about building a pipeline where each AI tool's output feeds the next, compounding the value at each step.

Q7: Track 6 (Governance) covers Responsible AI. In the context of this UCC risk platform, why is Responsible AI specifically critical?

Risk scores directly influence credit decisions — a biased model could illegally discriminate against borrowers, triggering regulatory violations and adverse action reporting requirements
It is required for open-source licensing compliance when using Anthropic's Claude API
Governance tools are needed to track API usage costs and prevent runaway AI spending
Responsible AI is only relevant for consumer-facing products, not B2B commercial credit systems
Correct! Commercial credit decisions are regulated by the Equal Credit Opportunity Act (ECOA) and Fair Credit Reporting Act (FCRA). If an AI risk score leads to a loan denial, the lender must be able to explain why — and prove the model doesn't discriminate based on protected characteristics. A black-box risk model with no audit trail isn't just technically risky — it's a legal liability.
Not quite. Commercial credit systems using AI risk scores face specific legal requirements: ECOA and FCRA require adverse action notices, the EU AI Act classifies credit scoring as high-risk AI, and regulators expect explainability and bias audits. A risk score that can't be explained or audited is a compliance violation, not just a technical gap.

Summary & Next Steps

You now have the map. Here's what you know:

  • AI-SDLC = AI at every phase, not just code generation. Planning, design, implementation, testing, deployment, and governance all have AI tools in this series.
  • One domain, one platform. Every course builds a real component of the UCC Lien Risk platform — data pipeline, REST API, frontend dashboard, AI tools layer, and governance audit. Nothing is thrown away between courses.
  • 6 tracks, 22 courses. Foundation → Build (Data/API/Frontend) → AI Tooling → Agentic AI → Project Mgmt → Governance. Enter at your level; the domain and prompt engineering courses are universal prerequisites.
  • The AI flywheel: Jira AI → OpenSpec → Gemini → Tests → MCP → Agent → Governance. Each tool's output becomes the next tool's input. That chain is what makes AI-SDLC transformative, not just incremental.
  • GCP + Local: Every Build Track course runs fully locally with Docker + DuckDB. No cloud account required. Skills transfer 100%.
Where to Go Next

Your first course: US Public Records — Commercial Credit Data Reference. This is the domain intelligence backbone. It takes 45–60 minutes, has no code, and will make every subsequent course 3x faster to follow.

Your second course: Prompt Engineering Masterclass. The RCTF framework, CLAUDE.md context engineering, and agent prompt design will make every AI tool in this series significantly more effective from day one.

Then: Pick your path above and follow the track that matches your role and goals. Every course will link forward to the next one.