back to work CASE 01 · 2026

PRODUCT DESIGN + AI BUILD · TWO-SIDED MARKETPLACE

marketplace poc

From an empty repo to a live marketplace in 3 weeks.

I designed, built, and shipped a functional two-sided recruiting marketplace — solo — and let real recruiters and real roles answer the questions before core engineering wrote a line.

{{ s.n }}
{{ s.label }}
{{ m.label }}
{{ m.value }}
THE ONE-LINER

Within 13 days of launch, 9 competing recruiters submitted 60 real candidates across 8 live roles. The PoC closed its first placement — a hire working at the company today, validating the entire marketplace loop end to end and becoming the blueprint for the version now built into the core platform.

CONTEXT

The company — a B2B recruiting platform — was betting its future on a marketplace model: multiple freelance recruiters competing to fill the same roles. Building that directly into the core platform meant months of engineering against unvalidated assumptions.

My proposal: don't. Build a standalone PoC first, put real recruiters and real roles in it, and let live behavior answer the questions before core engineering starts.

The stakes were real by design — the roles were live openings, the company's own hiring across engineering, product, sales, and marketing, plus client roles run through an account manager. We weren't simulating a marketplace; we were betting our own hires on it. I owned the whole thing: product strategy with the CEO, UX, UI, frontend and backend build (AI-assisted), deployment, and instrumentation.

THE FOUR QUESTIONS WORTH 3 WEEKS
Q{{ q.no }}

{{ q.title }}

{{ q.body }}

THE DECISION THAT DE-RISKED EVERYTHING

The riskiest assumption was competition sensitivity. My answer was structural, not cosmetic.

THE CORE DECISION

Private pipelines.
Every recruiter sees only their own candidates — competition exists, but is never visible friction. This single choice, made in week 1, is why activation never became the problem everyone feared.

THREE WEEKS, IN FAST-FORWARD
{{ w.label }}

{{ w.body }}

{{ s.n }}
{{ s.label }}
WHAT 13 DAYS OF LIVE DATA SAID
{{ d.n }}

{{ d.label }}

Supply — the thing everyone worried about — turned out to be the healthy part.

THE REFRAME — WHERE THE POC EARNED ITS KEEP

In the first 13 days, 45% of candidates were rejected — mostly at the first stage. Reject rates swung wildly between roles: 9% on the best-calibrated role vs. 75% on the worst.

"Recruiters were sending profiles that had nothing to do with the role — and nobody had given them a way to ask."

— a hiring manager, in a production note

Not a sourcing-quality gap — a calibration gap. The bottleneck isn't recruiter activation; it's pipeline progression. That reframing redirected what the core platform prioritized — before a single line of core code was written.

BUILD WITH USERS, NOT FOR THEM

Because the loop was live and small, feedback arrived in days. Gaps surfaced in production → features shipped inside the PoC window.

LIVE USAGE SURFACEDWE SHIPPED
{{ r.problem }} {{ r.fix }}
OUTCOME

The loop closed all the way. One completed placement — sourced by a marketplace recruiter, progressed through screening, interviews and offer, and hired. An engineer at the company today.

{{ o.kind }}

{{ o.text }}

WHAT HAPPENED NEXT?

The experiment answered the questions that mattered: recruiters participated, candidates were submitted, and the marketplace model worked.

The insights gathered from real usage reshaped product priorities and became the blueprint for the marketplace now running inside the platform.

Shipped in the core platform today

Confidential case study

The details are protected. You'll find the password in my application, or reach out and I'll share access.

Incorrect password — try again.

Request access →

NEXT PROJECT

antibody id

Designing a scalable, compliant platform for safer transfusion diagnostics.

← all work get in touch ↗