JD Case Study · PE portfolio
Applied AI Engineer — Forward-Deployed across Portfolio
A forward-deployed Applied AI Engineer embedded across a leading private-equity firm’s portfolio (undisclosed) — parachute into a company, find the highest-leverage AI opportunity, build it full-stack, productize it into a reusable template, train the team, and leave working software instead of slides. This is the stretch goal on my board and the one I want most. I’ll be straight: full-stack at staff-engineer depth is the bar I’m reaching for. But the AI-native half of this role — Claude Code, MCP, custom skills, context engineering, tool benchmarking, enablement — isn’t aspirational for me. It’s what I already do every day, and this site is the receipt.
What they’re actually buying
Strip the title down and the firm is buying compounding AI leverage across a portfolio from one operator. Not a contractor who ships a prototype and leaves; a builder who finds the step-change opportunity, ships it to production, turns it into a template the next portfolio company can redeploy, and levels up every team they touch. The economic logic is reuse: a forward-deployed engineer only pays for themselves if they "leave systems, not slides." That instinct — build once, redeploy many — is the exact thing this entire portfolio is built on.
How I’d own it — first 90 days
Days 1–30
Embed and find the lever
Drop into the first portfolio company with commit access. Map the product surface, the legacy architecture, and the customer workflows. Partner with the CPO/CTO to name the single highest-leverage AI opportunity — the one worth shipping in weeks, not quarters.
Days 31–60
Prototype in days, ship to prod
Build a working proof-of-concept fast — AI assistant, workflow automation, or an intelligent feature — then take it to production-grade with tests, monitoring, and docs. Pair with the in-house engineers so the knowledge transfers, not just the code.
Days 61–90
Productize and enable
Turn the solution into a reusable template — an MCP server, a skill, a documented playbook — deployable across the portfolio. Run an enablement session on AI-native dev workflows (Cursor / Claude Code) so the team is more capable than when I arrived, and line up the next engagement.
Signature analysis: the reusable-leverage funnel
Where a forward-deployed engineer actually pays off: not the prototype — the redeploy. One builder shipping one feature is a contractor. One builder who turns that feature into a template the next four portfolio companies reuse is a force multiplier. The prize is the bottom bar, and reaching it is a reuse discipline, not a coding-speed one. That’s the instinct I’ve been practising in public.
Illustrative portfolio funnel — the forward-deployed model’s core economics. The same build-once-redeploy-many instinct this whole site runs on: one battlecard engine reskinned per company, one design system ported per app, one skill library across tools.
Requirement → proof
| What the JD asks for | What I’ve already shipped |
|---|---|
| AI-native developer · Cursor / Claude Code power user (architect, debug, refactor, ship 10×) | This entire site and my operating tools are Claude Code / agent-built. AI isn’t autocomplete in my workflow — it’s how I architect, refactor, and ship. |
| MCP & extensibility — custom skills, tools, MCP servers | I author and run custom skills and MCP servers daily; my working toolkit is a live skill library, not a slide about one. |
| Context engineering — beyond RAG, persistent memory, deterministic output | A file-based memory + operating-doc harness gives my agents durable context and consistent behavior across sessions — context engineering as a daily practice, not a buzzword. |
| Tool scout — structured evals, adopt/skip recommendations | I run model/tool benchmarks with latency, cost, and LLM-judge quality scoring — adopt-or-skip calls backed by data, not vibes. |
| Productize patterns · train & enable teams | I turn solutions into reusable templates: this case-study engine, a ported design system, a skill library — built once, redeployed many. That’s the spine of the whole portfolio. |
| Product thinker — ARR/NRR, LTV/CAC, opportunity → roadmap | Years translating customer workflows into roadmaps and unit economics. LTV/CAC and funnel logic are my native tongue, with the SQL underneath. |
| Full-stack production builder, frontend → infrastructure | The honest stretch: I ship full products with AI — this site (Astro / React / TS), Supabase-backed apps — but I’m an analytics leader who builds, not a career staff SWE. I’d be the fastest-ramping non-traditional hire you meet, and I’d say so on day one. |
The same engine
The engine behind this whole portfolio is reusable leverage — one battlecard system reskinned per company, one design system ported per app, one skill library deployed across tools. That is precisely what a forward-deployed engineer monetizes: build it once, redeploy it across the portfolio, leave systems not slides. I’ve been running this play on my own work; this role just points it at someone else’s portfolio.
See the engine I reuse →This is the stretch goal on my board, and the one I want most. The full-stack-at-staff-level bar is real and I won’t paper over it — but “AI-native builder who ships reusable systems and levels up the teams around him” isn’t a goal for me, it’s a Tuesday. Point me at a portfolio company. — Paul Brown