AI-Assisted Career Curation
Using Claude Code to collect fifteen years of career artifacts, curate them through a bronze-silver-gold pipeline, and catch AI errors before they became public.
The Problem
I had fifteen years of career artifacts scattered across formats — annual reviews, accomplishments docs, presentations, old resumes, bios, project descriptions. Over forty files in total. Some lived in Google Docs, some in Confluence, some in folders I hadn't opened in years. Individually, each one captured a moment. Together, they told a story I'd never actually assembled.
The obvious move: feed everything to an AI and let it synthesize a portfolio. That doesn't work. Not because the AI can't write — it writes fluently, confidently, and fast. The problem is that confidence doesn't correlate with accuracy. An LLM will blend attributions across companies, hallucinate specific tools you never used, and present fabricated details with the same tone as verified facts. When the output represents your professional reputation, "close enough" isn't acceptable.
The real question wasn't whether AI could help — it was how to design a process that captured AI's speed while maintaining the accuracy standards of content that would be public, permanent, and attached to my name. That's a data quality problem, not a writing problem.
The Medallion Model: Bronze, Silver, Gold
I borrowed a concept from data engineering: the medallion architecture. Raw data enters at bronze, gets cleaned and structured at silver, and reaches publication quality at gold. Each tier has different rules, different quality gates, and different people responsible.
Bronze is raw capture. No editing, no curation, no quality judgment. I collected everything — forty-plus files including annual self-reviews, accomplishments documents, presentation decks, a consolidated resume, Q&A transcripts from clarifying sessions with Claude, and an AI-synthesized report from Atlassian's Rovo agent that crawled my Confluence history. The rule at bronze: capture everything, trust nothing.
Silver is themed narrative. Instead of organizing chronologically (which is how resumes work but not how portfolio content is consumed), I organized by theme: AI platform work, team transformation, product mindset, engineering practices, coaching philosophy, financial impact. Six themed files total, each curated through a structured process with Claude Code and verified by me. Silver is where AI errors get caught — or don't.
Gold is publish-ready. The case studies on this site, the resume page, the content you're reading right now. Gold has the strictest gates: metric verification, confidentiality review, and the question that matters most — would I stand behind this sentence in a job interview?
The per-theme process followed the same pattern each time: identify relevant bronze sources, run a clarifying Q&A session to fill gaps, check for confidentiality issues, write the curated narrative, then human review. That last step is where the model earns its keep — or reveals its limits.
Source Trust Hierarchy
The most transferable insight from this project isn't about portfolios — it's about working with AI on anything where accuracy matters. Not all source material is equally reliable, and AI doesn't know the difference. You have to.
I established a four-tier trust hierarchy:
- My direct Q&A answers — verbatim responses captured during clarifying sessions. Most reliable because I'm speaking from memory about my own experience, with the chance to correct myself in real time.
- My authored documents — annual reviews, accomplishments docs, things I wrote contemporaneously. Reliable but potentially incomplete — I wrote them for a specific audience, not for comprehensive accuracy.
- The consolidated resume — a bronze source, not ground truth. Resumes compress and simplify. Mine had at least one factual error that survived years of use (more on that below).
- AI-synthesized reports — useful for identifying gaps and suggesting structure, dangerous for specific claims. The Rovo synthesis introduced a hallucinated tool into the narrative. It appeared authoritative. It was wrong.
The principle: treat AI synthesis the way you'd treat a junior analyst's first draft. Useful for structure and coverage. Never trust the specifics without cross-referencing.
Four Errors Caught
This is the section that makes the process worth documenting. Every error below was caught by human review — not by the AI self-correcting, not by a second AI pass, not by automated checks. One review cycle, four catches.
The Micrometer Hallucination
The Rovo/Confluence synthesis report listed Micrometer as part of my team's observability stack. It's not. We use Datadog for infrastructure monitoring, Portkey and Opik for LLM-specific observability, and Pub/Sub into BigQuery for audit and lineage. Micrometer has never been part of the stack.
The hallucination propagated. It appeared in two separate silver-tier files before I caught it during review. The AI had no reason to doubt the source — it was in a document that looked authoritative — and I had no reason to doubt the AI until I read the output carefully enough to notice a tool name that didn't belong.
Remote-First Policy — Wrong Company
A silver draft attributed my remote-first meeting policy to Porch. I established that practice at Brinks, pre-COVID: if anyone on the call is remote, the meeting runs as fully remote. Porch was already fully remote when I joined — there was no policy to establish.
This is the sneakiest category of AI error: the right fact attributed to the wrong company. It reads correctly. It sounds plausible. It only fails if you were actually there.
Europe Contractors — Never Existed
A silver file described my global team as spanning the US, Latin America, India, and Europe. There are no Europe contractors. Never have been. The team is US-based employees, Latin American contractors, and Indian employees and contractors.
This error traced back to the consolidated resume itself — a bronze source, not an AI invention. The AI faithfully reproduced an error from my own document. The trust hierarchy matters in both directions: AI can amplify human errors just as easily as it creates its own.
LangGraph — Overclaimed
An early silver draft described LangGraph as a framework my team uses in production. The reality: another business unit ran LangGraph-based workflows through a vendor on infrastructure my team supports. My team facilitated the cross-BU collaboration. We haven't shipped LangGraph ourselves — it's under evaluation.
"Facilitated" and "managed" are different claims. "Enabled another team to use" and "adopted for production" are different stories. Precise language matters when the content represents your professional track record.
Four errors, one pass. Every one caught by a human reading carefully, not by any automated quality gate. That's the argument for keeping humans in the loop — not because AI is bad at writing, but because AI is bad at knowing when it's wrong.
The Conversation Is an Artifact
The most valuable content in this project didn't come from any document. It came from the Q&A exchanges — my verbatim answers to Claude's clarifying questions during the silver curation process. Those responses contain context, nuance, and reasoning that no annual review or accomplishments doc ever captured.
When Claude asked "what was the actual decision process for building Prediction Hub vs. buying VertexAI?" my answer included details about vendor lock-in concerns, observability gaps, and internal politics that I'd never written down anywhere. That answer, captured in a bronze Q&A file, became the foundation for a case study section. Without the intentional capture, it would have evaporated when the conversation window closed.
AI conversations are ephemeral by default. The insights, the reasoning, the back-and-forth that produced good output — all of it disappears unless you design for preservation. I built three mechanisms: Q&A transcript files in bronze (capturing the raw exchange), devlog entries during working sessions (capturing decisions and their rationale), and CLAUDE.md as persistent context across sessions (capturing conventions and hard-won lessons). Each serves a different purpose. Together, they mean the process of working with AI leaves a trail, not just the output.
The Meta-Layer
This project has a recursive quality worth naming. The infrastructure was scaffolded with Claude Code. The content was curated through a pipeline designed collaboratively with Claude. The documentation was written in working sessions with Claude. And this case study — about using AI to curate a career portfolio — is itself evidence of the skill it describes.
That's not clever circularity. It's the point. Knowing how to work with AI — when to trust output, when to verify, how to design quality gates, how to preserve context — is a distinct professional skill. It's the same set of instincts that govern production AI systems: observability (can you see what the AI did?), trust boundaries (which sources can you rely on?), human-in-the-loop design (where does automation stop and judgment begin?). The scale is different. The thinking is the same.
The medallion model, the source trust hierarchy, the four errors caught — none of these required building software. They required treating AI collaboration as a process design problem. That's the skill this case study demonstrates, and it's the skill that matters most as AI becomes a standard part of how knowledge work gets done.
Process Summary
| Phase | What | Quality Gate | |---|---|---| | Bronze | 40+ raw files collected | None — capture everything | | Silver | 6 themed narratives curated | Human review (4 errors caught) | | Gold | Case studies for kalmoe.com | Metric verification, confidentiality |