18 months of applied AI upskilling — agent architecture, multi-cloud orchestration, governance, and open-source tooling. Built in public, with AI tools wherever possible.
This is the actual working roadmap for Project Vigil — updated as things change, honest about what got dropped or delayed.
How This Roadmap Works
The goal of every term is to double-count ruthlessly: every class assignment feeds a self-imposed portfolio project or Waiven/ Vigil deliverable, and every build session reinforces what is being studied in class. I do not actively pursue certifications anymore, but there are listed if they align with the remaining activities as a potential to add more items to my resume, particulary if they are low cost or free. The class columns in the weekly tables show where that alignment exists and what the connection is. Traditional education has never felt like a great fit for me. I do well, but sometimes the outcomes don’t feel as meaningful as they could. Mark Twain said “I have never let schooling interfere with my education”. This is my attempt to use AI to stop schooling from getting in the way of getting educated.
Two factors sometimes override ideal coursework-to-project alignment, and both are called out explicitly:
- Cloud credit expirations. Azure credits expire first, which forces Azure-heavy work in Spring 2026 before the platform shifts to AWS in Summer. The build sequence follows the credits — not the syllabus.
- Order of operations. Some deliverables have hard dependencies: the Waive XXXXXXXXX engine requires model cards, which requires a working RAG pipeline. When those dependency chains land on lighter class weeks, the project work may run ahead of the classroom theory.
Where class content directly feeds a deliverable, it is noted in the table. Where it does not, both tracks are shown honestly so you can see what was competing for time that week.
Syllabi on hand through Summer 2026. Fall 2026 and Spring 2027 class columns are placeholders — they will be filled in when those schedules are confirmed.
Academic Structure
| Module | Title | Target |
|---|---|---|
| 1 | Landscape Report | Spring 2026, Week 8 ✅ |
| 2 | Governance Framework | Summer 2026 |
| 3 | Technical Prototype | Fall 2026 |
| 4 | Workforce Impact Assessment Tool | Fall–Spring |
| 5 | Student Roadmap | Spring 2027 |
| 6 | White Paper + Capstone | Spring 2027 |
Capstone deadline: May 1, 2027
Week Types
| Type | Meaning |
|---|---|
| SPRINT | Full build week — primary deliverable expected |
| LIGHT | Reduced sprint — coursework pressure week |
| RECOVERY | No sprint required — built in intentionally |
| FINALS | One post or commit only — protect the GPA |
| TRANSITION | Planning, admin, retrospective |
| BUILD | Free month — uninterrupted, no coursework |
Certifications
Already complete: Azure Copilot and Agent Administration Fundamentals
| Cert | Target | Status |
|---|---|---|
| AWS AI Practitioner (AIF-C01) | Summer S4 | Planned |
| Azure AI Engineer Associate (AI-102) | Summer S8 | Planned — prereq for AB-100 |
| AB-100 Agentic AI Business Solutions Architect | Fall F5 | Planned — expert level, requires AI-102 |
| AWS Agentic AI Demonstrated (microcredential) | Fall F9 | Planned |
| CEET Responsible AI | Fall F13 | Planned |
| AWS GenAI Developer Professional | Spring 2027 C5 | Planned |
Multi-cloud story: Azure depth (Copilot + Agent Admin + AI-102 + AB-100 expert-level) + AWS breadth (AI Practitioner → Agentic AI → GenAI Developer Professional) + vendor-neutral ethics (CEET). Six new certs plus two already complete by Spring 2027.
Open Source Tools
- Waive Builder — generates a portable
waive.mdcard. CLI live at projectvigil.org. Web app in progress. - Ledger Agent — drop folder watcher, builds efficiency ledger entries from URLs and files.
- PII Classifier — document processing pipeline with PII detection. Ships Spring W13.
The final project is Custos Bank, a fictitious financial institution undergoing an AI transformation, demonstrated as a platform integrating all portfolio projects.
Spring 2026 — Weeks 8–15
Machine Learning (AAI 201) · Ethics in AI (AAI 102) · Azure credits expire this term — burn first
| Week | Type | Machine Learning (AAI 201) | Ethics in AI (AAI 102) | Vigil Deliverable | Portfolio Build | Cloud | Tools | Cert Study |
|---|---|---|---|---|---|---|---|---|
| W8 ✅ | SPRINT | Midterm. Clustering & recommendation systems — direct input to model matching engine design. | M1: Bias & fairness foundations (Coeckelbergh). Ethical Analysis of AI assignment. | Ledger agent v1 shipped. Waive Builder scoped. | Azure OpenAI deployed. Azure AI Search configured. GitHub Pages + projectvigil.org live. | Azure | Azure OpenAI SDK, Azure AI Search, Python, GitHub Actions | AWS AI Practitioner: study begins |
| W9 | SPRINT | Q-Learning & OpenAI Gym. RL decision-making patterns apply to agent routing logic. | M2: Fairness in AI systems. Red Flag Review — audit an AI system for ethical risk. | Waive Builder v0.1 deployed. CLI open-sourced. Flask web app live on Azure. | RAG pipeline v0.1 — ingestion, chunking, embedding, retrieval. 50 docs in Azure AI Search. | Azure | LangChain, Azure AI Search SDK, Flask, Python OpenAI SDK | AWS AI Practitioner: continued study |
| W10 | SPRINT | Neural networks & generative AI fundamentals. | M3: Privacy & surveillance in AI. Ethical AI Audit — maps to governance framework. | Waive Builder v0.2 — free-text input with LLM PII filter. Web UX iteration. | RAG pipeline v0.2 — hybrid search. Chunking experiments measured. | Azure | LangChain, Azure AI Search (hybrid), ChromaDB, tiktoken | AWS AI Practitioner: practice exam |
| W11 | LIGHT | Deep learning architectures. CNN/transformer exploration. | M4: Accountability & transparency. Ethics by Design assignment — direct input to Vigil consent architecture. | Data Principles & Policy doc refined using Ethics coursework. Worker ownership language tightened. | Azure Document Intelligence setup. PII detection pipeline scoped. | Azure | Azure Document Intelligence SDK, Azure Content Safety | AWS AI Practitioner: practice exam |
| W12 | SPRINT | Fairness & model evaluation. Bias measurement techniques — applied to model card profiling and RAG retrieval quality. | M5: Human-centered AI. Ethical Design Plan — Vigil’s tiered consent architecture as case study. | Model Card Database v0.1 — first 3 model profiles (GPT-4o, Claude, Gemini). | PII Classifier architecture defined. Portfolio Project P1 “Industry Analyst” deployed on Azure AI Search. | Azure | Azure AI Search, Streamlit, ChromaDB, Azure Document Intelligence | AWS AI Practitioner: practice exam |
| W13 | LIGHT | Final project — Not Public. Double-count. | M6: Ethical frameworks. Vigil’s shared ownership model used as case study. | ML capstone = Not Public. Same document serves both. | PII Classifier v0.1. PDFs → structured data, PII detection. Azure Doc Intelligence. | Azure · GCP | Azure Document Intelligence, Vertex AI SDK (first activation), scikit-learn | — |
| W14 | FINALS | Final capstone submission. | M7: AI & social impact. Generative AI essay — framed around workforce displacement. | One commit. waive.md updated. | One commit. README updates only. | — | — | — |
| W15 | FINALS | Final reflection submission. | M8: Future of AI ethics. Coeckelbergh book review. Final course project. | Retrospective written. | Portfolio site scaffolded. | — | GitHub Pages, Jekyll | — |
Spring → Summer Break — 20 days
Summer 2026 — Weeks S1–S10
Python II (CIS 289) · Systems Analysis & Design (CIS 509) · AI-102 + AWS AI Practitioner cert targets
Multi-cloud mandate this term: Portfolio Project P3 is built on AWS Bedrock, not Azure. By end of Summer, the goal is to build an agent system on either platform without referencing documentation for every step.
| Week | Type | Python II (CIS 289) | Systems Analysis & Design (CIS 509) | Vigil Deliverable | Portfolio Build | Cloud | Tools | Cert Study |
|---|---|---|---|---|---|---|---|---|
| S1 | BUILD | OOP, decorators, function patterns — build identity card builder and agent components using Python classes. | SDLC concepts & systems thinking. Vigil product architecture used as real-world case study. | Waive Builder v1.0 — web app polished, projectvigil.org live. | LangChain + Semantic Kernel basics. First chains built. First model deploys on AWS Bedrock. | Azure · AWS | LangChain, Semantic Kernel, Boto3, Azure OpenAI SDK, Vertex AI SDK | AI-102 + AWS AI Practitioner: study begins |
| S2 | SPRINT | NumPy arrays & data structures — applied to benchmarking score matrices across platforms. | SDLC visual plan, feasibility study. Multi-cloud agent system framed as the system being designed. | Benchmarking Infrastructure v0.1 — identity cards scored against model API endpoints. | AWS: Bedrock Knowledge Bases, Bedrock Agents, Lambda. Single agent with tool use. Compared to Azure OpenAI function calling. | AWS | Boto3, AWS Lambda, Bedrock SDK, NumPy, promptfoo | AI-102: Computer Vision, Doc Intelligence. AWS AI Practitioner: Domains 1–2 |
| S3 | SPRINT | Pandas DataFrames & stats — model card data analysis and agent performance tracking. | Requirements gathering. Operations Assistant Agent requirements documented formally as class assignment. | Model Card Database v1 — 8 models profiled. Schema stable. | P3 “Operations Assistant Agent” — built on AWS Bedrock. Connects to SQLite DB + Bedrock Knowledge Base + email API. First full AWS build. | AWS | Boto3, Bedrock SDK, SQLite, Pandas, AWS Skill Builder | AI-102: Azure OpenAI Service, prompt engineering. AWS AI Practitioner: Domain 3 (Bedrock) |
| S4 | SPRINT | Data visualization, PyPlot — model matching scores visualized as radar charts. | Requirements modeling. Matching engine documented as a formal system for class assignment. | Model Matching Engine v0.1 — takes identity card input, scores against model card database. | Multi-agent system v0.1 — 3+ agents collaborating. Run on Azure OpenAI and AWS Bedrock. | Azure · AWS | AutoGen / CrewAI, Azure OpenAI SDK, Boto3, Matplotlib | AWS AI Practitioner: sit exam |
| S5 | LIGHT | Database integration & CRUD — identity cards and recommendation history stored in SQLite. | Project planning & scheduling. Full Vigil build sequence mapped as formal project plan. | Worker Model Recommendation v0.1 — card in, recommendation out. | Integration: Microsoft Graph API + AWS EventBridge, Step Functions, Lambda. Azure Logic Apps compared to AWS Step Functions. | Azure · AWS | MSAL (Graph API), Boto3, Azure Logic Apps, SQLite | AI-102: Azure Speech, Bot Framework |
| S6 | SPRINT | REST APIs — identity card submission and recommendation retrieval endpoints built as coursework. | Process modeling & decision tables. Mentorship pairing logic documented formally as class assignment. | Peer Mentorship Pairing Engine — design doc + matching logic v0.1. | P4 “AI Transformation Assessment Team.” Multi-agent, built on AWS Bedrock Agents. Cross-tested on Azure OpenAI. | AWS | Boto3 (Bedrock Agents), FastAPI, Azure OpenAI SDK, LangGraph | AI-102: intensive review |
| S7 | SPRINT | Web output & interactive charts — enterprise recommendation report dashboard built as coursework. | System architecture lab. Vigil’s three-layer architecture formally documented. Multi-cloud agent architecture diagrammed. Assignment = Vigil deliverable. | Enterprise Model Recommendation Report v0.1 — 5+ cards → department-level output. | P3 upgraded with enterprise integration. Same agent rebuilt in parallel on AWS. | Azure · AWS | Streamlit, Azure API Management, Boto3, Plotly | AI-102: mock exam |
| S8 | SPRINT | Threading & multiprocessing — applied to parallel model benchmarking across platforms. | Interface design challenge. Identity card UI designed and governance audit tool wireframes produced. | Workforce Displacement Framework — formalized. | AI-102 exam week. P5 “Production RAG System” scoped. Requirements drafted for Azure and AWS deployment. | Azure | Python threading, Azure SDK, Figma (wireframes) | AI-102: sit exam |
| S9 | LIGHT | Generators & data pipelines — applied to processing large evaluation datasets across platforms. | Testing & QA simulation. Test plans written for identity card builder and Production RAG System. | projectvigil.org v1 — card builder link, community hub, white paper placeholder live. | P5 “Production RAG System” v0.1. CI/CD pipeline. Azure AI Search + AWS Bedrock Knowledge Bases — both deployed. | Azure · AWS | Azure AI Search SDK, Boto3, GitHub Actions, Langfuse, pytest | AB-100: study begins |
| S10 | TRANSITION | Final project — identity card builder or ledger agent submitted as Python II deliverable. Double-count. | Final project — Vigil product architecture submitted as Systems Analysis deliverable. Double-count. | Summer retrospective. Deliverables audit. VIGIL_CONTEXT.md major update. | P1–P5 audited. Portfolio site updated. Multi-cloud cost ledger updated. | — | — | — |
Summer → Fall Break — 20 days
Fall 2026 — Weeks F1–F16
NLP · Data-Centric AI · AB-100 cert · Syllabi not yet on hand — class columns will be filled in when confirmed
| Week | Type | NLP | Data-Centric AI | Vigil Deliverable | Portfolio Build | Cloud | Tools | Cert Study |
|---|---|---|---|---|---|---|---|---|
| F1 | SPRINT | Syllabus pending | Syllabus pending | Workforce Impact Assessment Tool — requirements + scoring rubric v0.1. | P5 v1.0 complete. Production RAG on Azure + AWS with CI/CD + monitoring. | Azure · AWS | Azure ML SDK, Bedrock Guardrails, Langfuse, GitHub Actions | AB-100: agentic architecture |
| F2 | SPRINT | Syllabus pending | Syllabus pending | Workforce Impact Assessment Tool v0.1 — web app. Role + department → risk output. | P2 “M365 Adoption Analytics” v0.1 — Microsoft Graph API, Copilot usage by department. P6 “AI Governance Framework” — scoped. | Azure · AWS | Streamlit, MSAL (Graph API), Azure Responsible AI, Bedrock Guardrails | AB-100: tool use, agent memory |
| F3 | SPRINT | Syllabus pending | Syllabus pending | White Paper — Share First Methodology — first draft. | P6 v0.1 deployed. AI use case → risk assessment + mitigations + regulatory output. | Azure · AWS | LangGraph, Streamlit, Azure Content Safety, Bedrock Guardrails | AB-100: orchestration, routing |
| F4 | SPRINT | Syllabus pending | Syllabus pending | Model Matching Engine v1.0 — production. Individual + aggregate queries. | P6 v1.0 complete. Cross-platform governance audit tool. Published with cost comparison. | Azure · AWS | Semantic Kernel, Boto3, text-embedding models, FastAPI | AB-100: evaluation, safety |
| F5 | LIGHT | Syllabus pending | Syllabus pending | Tiered Observation Consent Architecture — v1 design doc. Three tiers. Worker approval flow. | P7 “AI Vendor Landscape Assessment” — research sprint. 15 platforms. Azure vs AWS vs GCP evaluated. | Azure · AWS · GCP | Azure Entra ID, AWS IAM + Cognito, GCP IAM, Vertex AI SDK | AB-100: sit exam · AWS Agentic AI: study begins |
| F6 | SPRINT | Syllabus pending | Syllabus pending | Enterprise Model Recommendation Report v1.0 — full pipeline live. | P7 draft complete. Category maps, evaluation matrices, 3-year cost projection. Multi-cloud framework. | AWS | Boto3, Lambda, S3, Vertex AI SDK, Plotly | AWS Agentic AI: Bedrock labs |
| F7 | SPRINT | Syllabus pending | Syllabus pending | Model Selection Protocol v1 — documented decision logic for model routing. | P7 v1.0 complete. Security: Azure Entra ID vs AWS IAM + Cognito. ADRs for P3–P5. | Azure · AWS | Azure Copilot Studio, Amazon Q, Boto3, MSAL | AWS Agentic AI: continued |
| F8 | RECOVERY | — | — | Review docs. No new builds. | Review all deployments. Update documentation. | — | Datadog, Langfuse (monitoring review) | Review weak areas |
| F9 | SPRINT | Syllabus pending | Syllabus pending | Workforce Impact Assessment Tool v1.0 — complete. Public URL. Demo-ready. | P8 Custos Bank: integration architecture scoped. All P1–P7 demo hooks designed. | Azure · AWS | GCP Cloud Run, AWS Lambda + Bedrock, Azure App Service, FastAPI | AWS Agentic AI Demonstrated: sit microcredential |
| F10 | SPRINT | Syllabus pending | Syllabus pending | Employee AI Adoption Toolkit v0.1 — card builder + assessment tool. | P2 “M365 Adoption Analytics” v1.0 complete. P8 Custos Bank: build begins. | Azure | MSAL, Azure Logic Apps, Copilot Studio, Streamlit | CEET: study begins |
| F11 | SPRINT | Syllabus pending | Syllabus pending | Benchmarking Infrastructure v1.0 — automated weekly runs. Results to ledger. | P8 Custos Bank: P1–P4 integrated. | Azure | Azure ML workspace, scheduled jobs, Weights & Biases, Langfuse | CEET: continued |
| F12 | LIGHT | Syllabus pending | Syllabus pending | White Paper v1.0 — complete draft. Share First Methodology with efficiency ledger data. | P8 Custos Bank: P5–P7 integrated. | Azure · AWS | Azure AI Search, Boto3, MLflow, Langfuse | CEET: intensive study |
| F13 | SPRINT | Syllabus pending | Syllabus pending | Vigil product architecture: all 17 deliverables documented. | P8 Custos Bank: full integration sprint. All 8 projects running together, demo-ready. | Azure · AWS | Full stack — all SDKs, GitHub Actions, Datadog | CEET: sit exam |
| F14 | SPRINT | Syllabus pending | Syllabus pending | Full integration sprint — card → matching engine → recommendation report. End-to-end documented. | P8 Custos Bank: polished and documented. | Azure · AWS | Full stack — all SDKs, pytest, Datadog | CEET: complete |
| F15 | LIGHT | Syllabus pending | Syllabus pending | Vigil product architecture final. All 17 deliverables mapped. | Portfolio site finalized. Interview narratives prepared. | — | GitHub Pages, Jekyll, Mermaid (architecture diagrams) | — |
| F16 | FINALS | Finals | Finals | VIGIL_CONTEXT.md updated. One commit. | One commit. Fall retrospective. | — | — | — |
Fall → Spring Break — 30 days
Spring 2027 — Weeks C1–C16
Computer Vision · Intro to Business · Capstone = Project Vigil · Syllabi not yet on hand — class columns will be filled in when confirmed
| Week | Type | Computer Vision | Intro to Business | Vigil Deliverable | Portfolio Build | Cloud | Tools | Cert Study |
|---|---|---|---|---|---|---|---|---|
| C1 | SPRINT | Syllabus pending | Syllabus pending | Capstone proposal submitted. Advisor aligned. | Workforce AI Readiness Assessment scoped. | — | All platforms stable — no new tooling | AWS GenAI Developer Professional: study begins |
| C2 | SPRINT | Syllabus pending | Syllabus pending | Peer Mentorship Pairing Engine v1.0 — live. | Readiness Assessment v0.1. Skills gap + reskilling program design. AWS Bedrock + Lambda. | AWS | Boto3, Lambda, Bedrock SDK, FastAPI | AWS GenAI Professional: Domain 1 |
| C3 | SPRINT | Syllabus pending | Syllabus pending | Capstone Ch 1–2: Introduction and Literature Review. | Ch 1–2 also serves as transformation methodology literature review. | — | Research tools, Gemini Deep Research, Claude Code | AWS GenAI Professional: Domain 2 |
| C4 | SPRINT | Syllabus pending | Syllabus pending | Capstone Ch 3–4: Methodology and System Design. Three-layer architecture documented. | P11: Full AI Transformation Proposal — assembly begins. All 8 components live. | Azure · AWS | Full stack — Azure + AWS SDKs, LangGraph, Semantic Kernel | AWS GenAI Professional: Domains 3–4 |
| C5 | SPRINT | Syllabus pending | Syllabus pending | Capstone Ch 5–6: Results and Discussion. 18 months of efficiency ledger data. | P11 complete. Financial model populated with 18 months of real cloud spend data. | Azure · AWS | Full stack, Weights & Biases, MLflow, Langfuse | AWS GenAI Developer Professional: sit exam |
| C6 | SPRINT | Syllabus pending | Syllabus pending | Student Roadmap published on projectvigil.org. Free forever. | Specialization project v1.0 complete. Portfolio site updated. | — | GitHub Pages, Jekyll, Claude Code | AWS GenAI Developer Professional: complete |
| C7 | SPRINT | Syllabus pending | Syllabus pending | Capstone final draft complete. | P11 complete. Assessment → RAG → governance → vendor brief → analytics all connected. | — | Full stack — all platforms and tooling | — |
| C8 | SPRINT | Syllabus pending | Syllabus pending | Capstone submitted. Job search launched. 5 target applications sent. | Apply strategically. Target mid-market companies early in AI journey. | — | LinkedIn, portfolio site, demo environments live | — |
| C9 | LIGHT | Syllabus pending | Syllabus pending | Portfolio walkthrough. 3-minute pitch rehearsed. | 10 informational interviews targeted. | — | Demo environments, Loom (walkthroughs) | — |
| C10–C14 | LIGHT | Syllabus pending | Syllabus pending | Ongoing applications. Vigil continues as a living product regardless of outcome. | Catch up, iterate, or rest. Keep applying. | — | — | — |
| C15 | FINALS | Finals | Finals | One commit. One post. | — | — | — | — |
| C16 | FINALS | Finals | Finals | Done. Final context update. Archive everything. | 8 projects, 8 certifications, 3 cloud platforms, 18 months documented. | — | — | — |
Multi-Cloud AI Progression
| Skill | Azure | AWS | GCP | Tools & Platforms |
|---|---|---|---|---|
| LLM calls | Azure OpenAI Service | Amazon Bedrock | Vertex AI Gemini | Claude Code, Cursor, Ollama (local), Anthropic API, Perplexity |
| RAG | Azure AI Search | Bedrock Knowledge Bases | Vertex AI Search | LangChain, LlamaIndex, ChromaDB, promptfoo |
| Agents | Semantic Kernel, Copilot Studio | Bedrock Agents, Step Functions | Vertex AI Agents | LangGraph, AutoGen / CrewAI, Claude Code (agentic) |
| Document AI | Document Intelligence | Textract | Document AI | Unstructured.io, Marker |
| Responsible AI | Content Safety | Bedrock Guardrails | Vertex AI safety | promptfoo (eval), CEET framework, Langfuse |
| MLOps | Azure ML workspace | SageMaker, CodePipeline | Vertex AI Pipelines | Weights & Biases, MLflow, GitHub Actions, Datadog |
| Identity / Auth | Entra ID | IAM, Cognito | Cloud IAM | MSAL libraries, 1Password (secrets) |
| Automation | Logic Apps, Power Automate | Step Functions, EventBridge, Lambda | Cloud Functions | GitHub Actions, Datadog (monitoring) |
Last updated: March 20, 2026 — v2.4. Tools column added to all term tables showing frameworks, SDKs, and platforms used or learned each week.