Method · Career Search

Career search with Claude.

A four-phase methodology I built running my own search through 2025 and 2026, before Facet existed as a product. The structure is deliberate: each phase's output feeds the next, so by the time you walk into a specific interview, you're working off compound context, not a fresh prompt. The kit at the bottom is the tools I actually used. Take the parts that fit. Throw the rest out.

Overview Phase 1 Phase 2 Phase 3 Phase 4 Tools Downloads
Structured context beats raw intelligence
Most engineers prep the same way: skim the resume, grind some LeetCode, Google a few behavioral questions, walk in. The problem is not effort. The problem is structure. Generic preparation produces generic answers, and generic answers do not survive a specific conversation with a specific panel about a specific role.

I ran a different process. Four phases, each one feeding the next, run in a Claude Project across about a week of focused iteration. The output was an interactive study guide I opened before every interview, a research doc with twenty-plus target companies, and a repeatable per-listing analysis I could run in an hour.

This page documents the methodology. The kit at the bottom is the artifacts that came out of it. Facet is the product I eventually built so I could stop recreating the manual setup every time.

Pipeline architecture
📄 Resume
📋 Tech docs
📂 Project READMEs
1 Candidate Profiling
Archetype 2-3 angles Strength/gap analysis
archetype + angles + gaps ↓
2 Market Research
Deep research Interview formats Stack overlap
20+ targets + formats ↓
3 General Study Guide
Variant scripts Deep dives Interactive prep page
reusable prep artifacts ↓
4 Per-Listing Analysis
Fit map Gap framing Company-specific tips
🎯 Interview ready

The core insight

An LLM with your resume can give you generic interview advice. An LLM with your resume, technical documentation, candidate profile, market research, and gap analysis can give you a prepared answer for the exact question a specific company is likely to ask about the exact gap in your profile for their exact role.

The compound context is the product. Each phase builds on the last.

Candidate profiling
Before you can prepare for interviews, you have to understand what you're selling. Most engineers skip this and jump straight to "help me prep," which is why they end up with generic advice for a specific situation.
1 Feed everything you have
Resume, technical references, architecture decision records, project docs, design docs. The more context, the sharper the output. I provided around 170 pages of technical references across four platforms. The technical docs were the highest-signal input by a wide margin. They contained architecture decisions, performance benchmarks, and tradeoff discussions a resume cannot capture.
1 Identify your archetype
Not your title. The pattern of how you work. Mine came back as "Builder": the engineer who gets dropped into a situation where something needs to exist and doesn't, then ships it end-to-end. That pattern held across both companies. It is a more useful framing than "Senior Platform Engineer" because it tells a story about how I operate, not just what my title was.
1 Define 2-3 angles
Different ways to frame the same experience depending on who you're talking to.
Angle Headline Best for
Security platform "I build security infrastructure: edge sensors, fleet management, threat intelligence" WAF / AppSec, security startups, detection engineering
Platform / DevEx "I treat infrastructure as a product: build systems, developer tooling, self-service platforms" DevTools, IDP, observability, platform teams
SRE / Infrastructure "I find the bottleneck, build the system that removes it, then hand it off" Infrastructure, reliability, cloud platforms
Output
Archetype identification
Output
2-3 positioning angles
Output
Strength/gap analysis with framings
Market research
Identify target companies where your profile has a genuine competitive advantage. Not just companies that are hiring, but companies whose interview process and culture reward what you specifically bring.
2 Two research passes
Pass 1: builder-friendly interviews. Companies with take-home assessments, paid work trials, portfolio reviews. If your strength is demonstrable output, optimize for formats that let you demonstrate it.

Pass 2: stack and domain overlap. Companies where your specific technical experience maps directly to their product or infrastructure.
2 Per-company intelligence
For each target: specific open roles, documented interview process, AI culture signals, stack overlap analysis, competitive advantage narrative, compensation range, and application tips specific to their process. The research identified twenty-plus targets ranked by signal convergence. Not just "who is hiring," but "where does my profile win."
Output
20+ ranked target companies
Output
Interview format analysis per company
Output
Competitive advantage narratives
General study guide
Build a reusable preparation artifact. Something you can open ten minutes before any interview and quickly navigate to the relevant talking points. This is the eighty percent that's company-agnostic.
3 Variant scripts, not single answers
"Tell me about yourself" gets three versions (Platform, Security, Builder), each leading with different experience and tuned to a different company type. You pick the variant based on who you're talking to. Each script is annotated with a "best for" callout naming specific companies from the research.
3 Navigable under pressure
Built as an interactive web page: grid of clickable tiles, floating header nav, modal system, keyboard shortcuts. Click a tile, get the modal with your talking points, arrow-key to the next section. Twenty-five-plus sections covering openers, project deep dives, behavioral variants, and a key numbers reference. Designed for the literal moment you're sitting in a waiting room.
Grid layout Modal system Variant tabs Keyboard nav Floating header
Output
Interactive study guide (HTML)
Output
Variant scripts per angle × question
Output
Project deep dives + key numbers
Per-listing analysis
For each specific job listing, produce a targeted analysis: how your profile maps to their requirements, where you're strong, where you're thin, and how to talk about the thin spots. The twenty percent that's company-specific, but the highest-signal twenty percent.
4 Fit map
Map each stated requirement to your actual experience. For each: the requirement, your relevant evidence, fit level (Strong / Partial / Gap), and if it's a gap, prepared framing that's honest but redirects to adjacent strength. "You don't have Go, here's how to handle it" is more useful than pretending the gap doesn't exist.
4 Interview tips and company hooks
Likely questions based on the gaps in your profile for this specific role. Interview tips tuned to their known process: take-home strategy, pair programming approach, system design framing. Company hooks: specific things to mention that show you've done homework on this company. Not "I like your mission." Specific technical decisions they've made that you have an informed opinion on.
Output
Fit map with gap framings
Output
Predicted hard questions + answers
Output
Company-specific cheat sheet
Tools and meta
The original methodology was built in a Claude Project over about a week of focused iteration. The Project layer is what makes the compound context possible: each conversation builds on what came before instead of starting from zero.

Claude Projects

Persistent context across conversations. Resume, tech docs, research outputs all in one project. Each conversation builds on the accumulated context from previous phases.

Deep Research

Long-running research tasks with citations for Phase 2. Two separate passes identifying twenty-plus target companies with interview format analysis.

Artifacts

Interactive study guide built and iterated in-conversation. HTML, CSS, and JS generated, reviewed, and refined across multiple sessions. Grid layout, modals, variant tabs.

The meta

Using AI to prepare for interviews about AI-augmented development. This page, the study guide, and Facet itself are all artifacts of the same methodology being described.

Total time: about a week (manual)

The original deliverables: an interactive study guide opened before every interview, a research doc with twenty-plus target companies, a repeatable process for per-company analysis, and this writeup. The study guide ended up being the highest-ROI artifact. I opened it before every panel.

Then I built Facet so the loop could keep running without recreating the manual setup for every search. Same methodology. Persistent context. Open source.

Downloads
Standalone tools from the methodology. Use them in a Claude Project or on their own. No account needed. The interview prep skill works with any agent that supports skills; everything else is HTML you open in a browser.
What you get. Click any thumbnail to expand.

Facet is this. On steroids. Lots of them.

Identity model that persists across every search, pipeline with rounds and snapshots, prep decks that survive across weeks, per-interviewer research, drift detection, the works. Same methodology. Without the manual setup every time.