praeducer
diff --git a/‎.claude/plans/README.md‎
Lines changed: 91 additions & 0 deletions b/‎.claude/plans/README.md‎
Lines changed: 91 additions & 0 deletions
diff --git a/‎.claude/plans/autonomize-transition-agent-handoff.md‎
Lines changed: 148 additions & 0 deletions b/‎.claude/plans/autonomize-transition-agent-handoff.md‎
Lines changed: 148 additions & 0 deletions
@@ -0,0 +1,91 @@
+# `.claude/plans/` — Session Entry Point
+
+**If you are a fresh Claude Code agent starting in this repo, read this file first.**
+
+## Current State (2026-04-11)
+
+**Branch:** `uat/mega-merge-apr-2026` | **PR:** [#39](https://github.com/praeducer/paulprae-com/pull/39) (draft, CI green)
+**All 7 feature branches merged.** 511 tests pass, resume quality 415, Autonomize AI is current employer everywhere.
+
+**Paul starts at Autonomize AI:** Monday April 13, 2026
+**Next step:** Paul runs UAT on Vercel preview, squash-merges PR #39 to main
+
+## Plan Documents
+
+### Active
+
+| File                               | Purpose                                           | Status                           |
+| ---------------------------------- | ------------------------------------------------- | -------------------------------- |
+| `remaining-phases-ssot.md`         | Writing-rules SSOT refactor roadmap (Phase A1-A9) | Phases A1+A3+A5 executing on UAT |
+| `content-quality-system-design.md` | 5-layer quality stack architecture                | Reference (locked)               |
+| `backlog.md`                       | Post-merge coding/automation backlog              | Active, reconciled 2026-04-11    |
+
+### Reference (human action needed)
+
+| File                               | Purpose                                      |
+| ---------------------------------- | -------------------------------------------- |
+| `human-tasks.md`                   | Manual tasks (DNS, SEO, mobile testing)      |
+| `production-monitoring.md`         | Post-deploy monitoring playbook              |
+| `production-qa-plan.md`            | Stakeholder-centered QA framework            |
+| `data-model-and-knowledge-base.md` | Phase 3 knowledge graph design (deferred)    |
+| `hotfix-multi-resume-bug.md`       | Multi-resume chat bug (tracked as issue #41) |
+
+### Completed (historical reference)
+
+| File                                     | Purpose                                             | Completed                           |
+| ---------------------------------------- | --------------------------------------------------- | ----------------------------------- |
+| `mega-merge-strategy.md`                 | 10-phase merge of 7 branches into UAT               | PR #39 open                         |
+| `mega-merge-review-prompt.md`            | Copilot SWE Agent review (30 issues, all addressed) | Issues in v2 plan                   |
+| `autonomize-transition-agent-handoff.md` | Career transition tooling refactor                  | PR #38 absorbed into UAT            |
+| `autonomize-transition-human-runbook.md` | Paul's pre-merge QA guide                           | Reference for future career changes |
+| `generic-jingling-mccarthy.md`           | Mega-merge execution plan (this session)            | Execution complete                  |
+
+## Authoritative Career Timeline
+
+Pinned in `tests/data-consistency.test.ts`. **Do NOT trust memory files for dates.**
+
+| Role                       | Company             | Start    | End       | Type                |
+| -------------------------- | ------------------- | -------- | --------- | ------------------- |
+| Solutions Architect        | **Autonomize AI**   | Apr 2026 | _current_ | full-time           |
+| Staff AI DataOps Engineer  | Arine               | Sep 2025 | Mar 2026  | full-time           |
+| Chief AI Architect         | Booz Allen Hamilton | Jul 2024 | Mar 2025  | full-time           |
+| Chief AI Officer, Founder  | Hyperbloom          | Jun 2021 | Aug 2025  | self-employed       |
+| Neuroinformatics Architect | TReNDS Center       | Jan 2022 | Sep 2023  | full-time           |
+| Enterprise AI Architect    | Amazon Web Services | Aug 2018 | May 2021  | full-time           |
+| Senior AI Architect        | Decooda             | Feb 2018 | Jul 2018  | full-time           |
+| Senior AI Engineer         | NeuroLex Labs       | Feb 2018 | Jul 2018  | part-time moonlight |
+| Analytics Consultant       | Slalom Consulting   | Jul 2015 | Jan 2018  | full-time           |
+
+## Quick Sanity Check
+
+```bash
+python3 -c "
+import json
+with open('data/generated/career-data.json') as f:
+    data = json.load(f)
+targets = {'Autonomize AI':'2026-04→None', 'Arine':'2025-09→2026-03',
+           'Hyperbloom':'2021-06→2025-08', 'NeuroLex Labs':'2018-02→2018-07',
+           'Decooda':'2018-02→2018-07'}
+for p in data['positions']:
+    if p['company'] in targets:
+        actual = f\"{p['startDate']}→{p['endDate']}\"
+        expected = targets[p['company']]
+        status = '✅' if actual == expected else '❌'
+        print(f'{status} {p[\"company\"]:20} {actual} (expected {expected})')
+"
+```
+
+## NVIDIA Tailored Content
+
+- Resume: `data/generated/tailored/Paul-Prae-Resume-NVIDIA.md` (95% grader score)
+- Cover letter: `data/generated/tailored/Paul-Prae-Cover-Letter-NVIDIA.md` (92% grader score)
+- Iterate: `npm run generate:tailored -- nvidia --force` then `npm run grade`
+
+## Tracking Issues
+
+- [#40](https://github.com/praeducer/paulprae-com/issues/40) — Phase A SSOT refactor
+- [#41](https://github.com/praeducer/paulprae-com/issues/41) — Multi-resume chat bug
+
+---
+
+_Last updated: 2026-04-11 by Claude Opus 4.6 on `uat/mega-merge-apr-2026`_
@@ -0,0 +1,148 @@
+# Autonomize Transition — Agent Handoff
+
+**Purpose:** Full context for a Claude Code agent picking up this work in a new session. Read top to bottom before touching code.
+
+**Branch:** `feat/autonomize-ai-career-update`
+**Parent:** `main`
+**Target merge date:** Sunday April 12, 2026
+**Paul's new role starts:** Monday April 13, 2026
+
+## Goal
+
+Bundle two things in one PR:
+
+1. **Phase 2 (tooling):** refactor the career-data pipeline so that hardcoded current-employer strings no longer exist in prompts, hero copy, quick actions, or quality checks. Current role must be derived at build time from `data/generated/career-data.json`.
+2. **Phase 1 (data):** end-date Arine at March 2026, add Autonomize AI as Solutions Architect starting April 2026, run the full pipeline, and commit all regenerated outputs.
+
+Phase 2 comes first. Phase 1 is then the test that proves the refactor works: after Phase 2, Phase 1 should touch only data files (`Positions.csv`, `companies.json`, `positions.json`) and regenerated artifacts.
+
+## Critical context discovered during Phase 1 research
+
+- **Structured position data comes from `data/sources/linkedin/Positions.csv`**, NOT from `data/sources/knowledge/career/positions.json`. The JSON file is wrapped as a Claude knowledge-base context entry via `lib/ingest/knowledge.ts`, not parsed into the structured `data.positions` array. See `lib/ingest/normalizers.ts::normalizePositions()` — it reads LinkedIn CSV rows only.
+- **`Positions.csv` is gitignored** per CLAUDE.md. The generated `career-data.json` IS committed, so Vercel builds work without the CSV.
+- **`career-data.json` position shape is camelCase** (`startDate`, `endDate`, no `isCurrent` field). A role is "current" iff `endDate === null`. See `lib/types.ts::CareerPosition`.
+- **`positions.json` in the knowledge base was stale** (wrong start dates, stale `is_current` flags for Hyperbloom/Modular Earth that conflict with the CSV). It's a knowledge-base artifact, not a source of truth. Updating it is "nice to have" for consistency but doesn't drive behavior.
+- **`build-prompts.ts` runs `buildSystemPrompt(mode)` from `lib/agent/context.ts`**, which already has a placeholder-substitution mechanism. Extending it is the right place to inject `{{CURRENT_ROLE_SENTENCE}}`.
+- **`lib/constants.ts` must stay free of server-only imports** (fs, path) — client components import from it. Therefore HERO_DESCRIPTION cannot read career-data.json directly; it must import from a pre-generated `lib/generated/current-role.ts`.
+- **Three perpetual side-ventures have `endDate === null`** alongside any real current employer: Modular Earth, Hyperbloom (but wait — the CSV now end-dates Hyperbloom at Sep 2025), and Paul Prae (self-brand rows). The `getCurrentRole()` heuristic in `lib/career-data.ts` excludes these via a `SIDE_VENTURE_COMPANIES` set and picks the latest-startDate active role.
+- **career-objectives.json** (`data/sources/knowledge/strategy/career-objectives.json`) is PRIVATE and explicitly excluded from the knowledge base via `lib/ingest/knowledge.ts::EXCLUDED_FILES`. Do NOT load it into prompts.
+
+## Files touched by Phase 2 tooling refactor
+
+- `lib/career-data.ts` — added `getCurrentRole()`, `getCurrentEmployer()`, `formatCurrentRoleSentence()`, `formatCurrentRoleHero()`, and the `SIDE_VENTURE_COMPANIES` set.
+- `lib/generated/current-role.ts` — NEW file, auto-generated by `build-prompts.ts`. Exports client-safe `CURRENT_ROLE_TITLE`, `CURRENT_EMPLOYER`, `CURRENT_ROLE_SENTENCE`, `CURRENT_ROLE_HERO` string constants. Committed so prod builds on Vercel don't need to re-derive.
+- `scripts/build-prompts.ts` — extended to ALSO emit `lib/generated/current-role.ts` from career-data.json on every run.
+- `lib/constants.ts` — imports `CURRENT_ROLE_HERO` from `lib/generated/current-role`; `HERO_DESCRIPTION` composes it with the static prefix.
+- `lib/agent/context.ts` — substitutes `{{CURRENT_ROLE_SENTENCE}}` in prompt templates.
+- `lib/prompts/career-chat.few-shot.md` — replaced the hardcoded "Currently he's a Staff AI DataOps Engineer at Arine…" sentence with `{{CURRENT_ROLE_SENTENCE}}` placeholder. Also updated past-tense references to Arine bullets to be role-neutral.
+- `app/components/QuickActions.tsx` — replaced the "at Arine" quick action with a role-agnostic "Recent work" phrasing.
+- `lib/resume-quality.ts` — added `"Autonomize AI"` to the `MAJOR_COMPANIES` array (kept Arine).
+
+## Files touched by Phase 1 data update
+
+- `data/sources/linkedin/Positions.csv` — Arine row `Finished On` → `Mar 2026`; new Autonomize AI row prepended.
+- `data/sources/knowledge/career/positions.json` — sync'd to match CSV (even though it doesn't drive structured data).
+- `data/sources/knowledge/career/companies.json` — new `autonomize-ai` entry, metrics deliberately empty (do not invent figures).
+- `CLAUDE.md` — brand voice line updated ("healthcare domain expertise (Autonomize AI, Arine, BCBS, Humana ecosystem)").
+- `lib/prompts/resume-writer.system.md` — same brand voice list update.
+- `lib/prompts/resume-writer.few-shot.md` — same.
+- `docs/uat-checklist.md` — current-role assertion updated to Autonomize AI.
+- Regenerated & committed: `data/generated/career-data.json`, `data/generated/Paul-Prae-Resume.md`, `data/generated/Paul-Prae-Resume.pdf`, `data/generated/Paul-Prae-Resume.docx`, `lib/generated/system-prompts.ts`, `lib/generated/current-role.ts`, `public/Paul-Prae-Resume.{md,pdf,docx}`.
+
+## Pipeline commands (run in order from repo root, WSL Ubuntu)
+
+```bash
+# Sanity: ensure branch and clean tree
+git status
+git log --oneline -5
+
+# Phase 2 tooling is committed BEFORE running pipeline, so the pipeline's
+# output reflects the new derivation.
+npm run ingest                # CSV → career-data.json
+npm run build:prompts         # career-data.json → lib/generated/system-prompts.ts + current-role.ts
+npm run check:quick           # fast JSON + hash validation
+npm run generate              # Claude Opus 4.6 → Paul-Prae-Resume.staging.md (needs ANTHROPIC_API_KEY)
+npm run approve               # staging → approved
+npm run export                # Pandoc + Typst → PDF/DOCX, sync to public/
+npm run check                 # full pre-push release checklist
+```
+
+If `ANTHROPIC_API_KEY` is missing: stop after `build:prompts`, commit source + intermediates, mark PR as WIP in description. Paul will run the AI generation step.
+
+If `npm run export` fails: check `which pandoc && which typst`. WSL Ubuntu has both (pandoc in /usr/bin, typst in ~/.local/bin/typst per verified-at-session-start).
+
+## QA expectations before opening PR
+
+- `npm run check` passes fully
+- `npm test` passes
+- `npm run build` succeeds
+- `localhost:3000` hero text mentions Autonomize AI
+- Chat answer to "Where do you work now?" cites Autonomize AI
+- Chat answer to "Tell me about your time at Arine" uses past tense with Sep 2025 – Mar 2026 dates
+- `data/generated/Paul-Prae-Resume.md` has Autonomize as first position with "Apr 2026 – Present"
+- `public/Paul-Prae-Resume.pdf` opens and renders correctly
+
+## Commit strategy
+
+Each logical unit is its own commit, and EVERY commit is pushed immediately to origin to protect against machine crash. Rough sequence:
+
+1. `docs: autonomize transition handoff docs` — the two files in `.claude/plans/`
+2. `feat: derive current role from career data (tooling)` — Phase 2 refactor only (no data change)
+3. `feat: add Autonomize AI position and end-date Arine` — source data edits
+4. `chore: regenerate career-data.json + system prompts` — pipeline intermediate outputs
+5. `feat: regenerate resume markdown, PDF, DOCX for Autonomize role` — resume artifacts
+6. `docs: update UAT checklist and brand voice for Autonomize transition` — misc doc updates
+
+Do NOT merge to main. The PR stays open for Paul to merge Sunday April 12.
+
+## STATUS: COMPLETE — PR #38 open and ready for review
+
+All 5 commits landed, pushed, and passing:
+
+1. ✅ `docs: autonomize transition handoff docs`
+2. ✅ `feat: derive current role from career data at build time` (Phase 2 tooling)
+3. ✅ `feat: add Autonomize AI and end-date Arine in career data` (Phase 1 data)
+4. ✅ `chore: regenerate career-data.json + system prompts for Autonomize`
+5. ✅ `feat: regenerate resume + exports for Autonomize AI role`
+
+PR: https://github.com/praeducer/paulprae-com/pull/38
+Tests: 493 passed, 0 failed
+Build: clean, zero TypeScript errors
+Resume quality: 423 (+5% over previous)
+
+Remaining manual QA (Paul to do before merge):
+
+- Chat: "Where do you work now?" → expect Autonomize AI
+- `/resume` page: Autonomize first position
+- PDF download: renders correctly
+- Professional summary: says "13+ years" (not 15)
+
+## Known gotchas
+
+- **`lib/constants.ts` is imported by client components.** Never import `fs` or `path` there. The generated `lib/generated/current-role.ts` is pure string constants — safe for client.
+- **Prompt cache keys.** `lib/generated/system-prompts.ts` strings must be byte-identical between local and Vercel. If you change the prompt templates, regenerate with `build:prompts` and commit.
+- **LinkedIn CSV is gitignored but used by ingest.** On a fresh checkout without the CSV, `npm run ingest` fails. Committed `career-data.json` is the fallback for CI and Vercel. Do NOT force-run `ingest` in CI.
+- **`npm run check` hashes inputs and outputs** (see `scripts/release-check.ts` and the `writeIngestHash` calls in ingest). If ingest runs but output is identical, that's fine; if output changes without input changes, something is wrong.
+- **The few-shot `{{CURRENT_ROLE_SENTENCE}}` placeholder must be resolved before the prompt is baked** (`build-prompts.ts` runs `buildSystemPrompt` which runs `context.ts` substitution). Verify by grepping the committed `lib/generated/system-prompts.ts` — it should contain "Autonomize AI", NOT `{{CURRENT_ROLE_SENTENCE}}`.
+
+## Out of scope (do NOT touch)
+
+- PR #36 — the Autonomize team intro deliverable. Separate PR, separate branch.
+- `docs/examples/tailored-resume-*.md` — 3 historical artifacts, leave them.
+- `data/generated/reference-library/resumes/Paul-Prae-Resume-20260305.md` — historical snapshot.
+- `tests/fixtures/sample-data.ts` and `tests/resume-parser.test.ts` — Arine is still in the resume, so assertions pass.
+- Two intervening roles between Arine and Autonomize — intentionally omitted per Paul's instruction.
+- Autonomize metrics (customer counts, funding) — Paul must provide; do not invent.
+- `data/sources/knowledge/strategy/career-objectives.json` — PRIVATE.
+
+## Paul's brand voice reminders (used for AI resume generation)
+
+From CLAUDE.md §Brand Voice:
+
+- Tone: confident, technically precise, action-oriented
+- Perspective: third-person professional (no "I")
+- Emphasis: AI engineering leadership, healthcare domain expertise, Fortune 500 delivery, full-stack capability
+- Quantify impact where data supports it
+- Target roles: Principal AI Engineer, Solutions Architect, Senior Engineering Manager
+- Avoid: buzzword stuffing, vague claims, passive voice, overly humble hedging
+- Length: ~2 pages rendered