AI Product · Designer & Creator

ReferLink ,
AI Job Hunting Platform

You've applied to 200 jobs. You've heard back from 3. The market is brutal, and the only thing that actually moves the needle is a referral from someone on the inside. ReferLink makes that possible at scale.

RoleDesigner & Product Creator
Team2 Engineers (Backend)
StatusIn development
ToolsCursor, Claude Code
ReferLink app

Designer-led, engineer-supported: I lead product vision, UX strategy, and all design decisions. Two backend engineers collaborate with me on infrastructure and data systems. This case study focuses on the product thinking and design rationale.

The job market is brutal. Applications go nowhere.

I experienced it firsthand, and so did everyone I talked to. You spend hours customizing resumes, writing cover letters, clicking "Easy Apply" hundreds of times. And you hear nothing back. The market is saturated, recruiters are overwhelmed, and your application is one of 500 in their inbox.

But when someone refers you? Your resume goes to the top of the pile. Referrals account for 30-50% of all hires, yet most job seekers don't have the right connections. The problem isn't effort. It's access.

What job seekers told me
"I've sent 200+ applications and gotten 3 responses"

"I know referrals work, but I don't know anyone at these companies"

"I feel weird cold-messaging strangers on LinkedIn for help"
What referrers told me
"I'd love to help people, I get a bonus too, but random LinkedIn DMs feel sketchy"

"If I could see their full profile and why they're a fit, I'd refer more people"

The deeper insight: both sides want the same outcome. Job seekers need introductions. Employees want to earn referral bonuses. The friction isn't motivation, it's trust and access.

The second insight: even with a referral, job searching still eats hours every day. Checking boards, tailoring resumes, tracking applications. The manual work never stops. I saw two problems that needed solving, in the right order.

Two problems. One sequence. The order matters.

You can't automate referral discovery without first having a referral network. So I designed a two-phase approach: build the infrastructure first, then add intelligence on top.

Phase 1
Referral Marketplace + Dashboard
Connect job seekers with employees who can refer. Give both sides a unified place to track everything. Solve the access problem first, create the infrastructure Phase 2 needs.
Phase 2
AI Agent (In Development)
Autonomous job discovery across 50+ sources. Semantic resume tailoring per job. Referral path finding. Application tracking. Runs 24/7, you check the dashboard, not the job boards.
Why this order
Infrastructure before intelligence
You can't automate referral discovery without first having a referral network to tap into. Phase 1 creates the infrastructure. Phase 2 adds the intelligence layer on top.
ReferLink marketplace - Find Referrers dashboard
1

Explore referrers: Browse employees at top companies willing to refer, filterable by role, company, and location

2

One-click request: "Request Referral" removes the friction of cold outreach. No awkward LinkedIn DMs needed

3

Profile completeness: Seekers build trust through verified LinkedIn, uploaded resume, and target role clarity

4

Role switching: Users toggle between Referrer and Job Seeker modes, both sides of the marketplace in one account

Stop networking. Start matching.

You shouldn't need to spend months building relationships just to get your resume seen. Employees earn referral bonuses and want to help. Job seekers need introductions. The platform removes the awkwardness and matches them directly.

For job seekers

  • Browse employees at target companies willing to refer
  • Send low-stakes introduction requests (limit on simultaneous requests reduces spam)
  • Track all referral conversations in one dashboard
  • Combine with application tracking, one product for the entire search

For referrers

  • See full candidate profile before agreeing to refer
  • Match score shows why this person fits their company
  • Control over who you help, no obligation to refer everyone
  • Referral bonus tracking built in

Your job search runs 24/7. You don't.

Even with referral access solved, the daily grind remains: scanning 50+ job boards, reading descriptions, tailoring resumes, tracking applications. That's hours of repetitive work every single day. The AI agent takes over.

What the agent does

  • Autonomous job discovery, continuously scrapes LinkedIn, Greenhouse, Lever, company career pages, detects opportunities within hours of publication
  • Semantic matching, alerts only to high-priority matches, not every posting
  • Resume tailoring, AI adapts resume language to match each job description automatically
  • Referral path finding, identifies employees at target companies who can refer
  • Application tracking, unified dashboard, status, and follow-ups
The stickiness insight
Referrals are episodic, you need one, then you're done until your next job search. The real challenge: how do you create daily engagement in a product people only need occasionally?

The AI agent solves stickiness. If it runs daily and surfaces opportunities, people check the dashboard daily. The sequencing isn't just technical, it's business model.
AI agent runs while you sleep
Overwhelmed by manual job search

If users can't see why, they won't trust it.

A lesson I carried from my LivePerson work: AI recommendations only work if users understand them. Every match in ReferLink shows a clear explanation: which skills matched, which experience gaps exist, and why this role was surfaced. No black boxes.

For candidates
See exactly why each job was matched. Control the search parameters. Review and approve before anything is sent.
For referrers
See candidate profile, match score, and why this person was selected for their company, before agreeing to refer.
For both
Human review at every critical step. AI does the legwork, humans make the decisions.

Building a marketplace teaches you things features can't

01
Retention is the real design problem

Referrals are episodic: you need one, then you're done. A two-sided marketplace forced me to think about daily engagement, not just functionality. That insight shaped the entire Phase 2 strategy.

02
Prototype to learn, build to ship

Lovable got me to a working prototype in days. Shipping the real product took weeks. I learned when to prototype (test unclear assumptions) versus when to just build (requirements are clear, execution matters).

03
Speed to learning beats speed to perfection

Putting something in front of users teaches you things interviews can't. Build, ship, learn, iterate. The product got meaningfully better after every round of real feedback.

Previous AI project Mind Flux Next AI project Tiny Closet