Semantic search and predictive AI that helps hiring teams find who fits — not just who matches keywords. I built the prototype solo, delivered it to the Lumeir founder, and led a team of 4 to scale it into the full product.

Lumeir — Recruit Smarter

Lead Designer · 6 months

AI B2B SaaS0→12 designers + 2 engineersv0 · Figma Make · Figma

Semantic search and predictive AI that helps hiring teams find who fits — not just who matches keywords. I built the prototype solo, delivered it to the Lumeir founder, and led a team of 4 to scale it into the full product.

40%
Candidate review time reduction
3 weeks
Solo prototype · founder sign-off
2 of 2
Founders signed off on final direction
Greenlit
Approved to build · queued post-beta
40%
Candidate review time reduction
3 weeks
Solo prototype · founder sign-off
2 of 2
Founders signed off on final direction
Greenlit
Approved to build · queued post-beta

Lead Designer · 6 months

AI B2B SaaS0→12 designers + 2 engineersv0 · Figma Make · Figma

Semantic search and predictive AI that helps hiring teams find who fits — not just who matches keywords. I built the prototype solo, delivered it to the Lumeir founder, and led a team of 4 to scale it into the full product.

Lumeir — Recruit Smarter

Semantic search and predictive AI that helps hiring teams find who fits - not just who matches keywords. I built the prototype solo, delivered it to the Lumeir founder, and led a team of 4 to scale it into the full product.

40%
Candidate review time reduction
3 weeks
Solo prototype · founder sign-off
2 of 2
Founders signed off on final direction
Greenlit
Approved to build · queued post-beta

Process — 6 months

The team discovered together. I decided where to go next.

Research was a team effort. Strategic decisions about where to go next was mine.

Discovery & Research

2.5 months · Team

Define & Vibecode

2 weeks · Solo

Refine Concept

2 weeks · Team

Lead Team Build

3 months · I Led

My ownership

My ownership

I built it first. Then I led the team to scale it.

I built it first. Then I led the team to scale it.

Hiring managers send requisition files. Recruiters manually re-read them every time they screen a candidate. I built an uploader that pulls requirements from the file automatically, filters candidates against it, and surfaces an AI analysis - so the decision is ready before the recruiter even opens a profile.

SoloFigma MakeUser testedFounder approved

The problem

Competitive positioning

The ATS was designed to filter keywords.
AI usage made the resume an unreliable signal.

Now, Lumeir wins where existing tools have no answer.

Resume signal reliability

AI adoption risesReliableUnreliable

Once resumes could be optimized in seconds, the signal recruiters relied on for years collapsed almost overnight.

Once resumes could be optimized in seconds, the signal recruiters relied on for years collapsed almost overnight.

The problem

The problem

Existing process: Not efficient. Not accurate. Not aligned.

The ATS was designed to filter keywords.
AI usage made the resume an unreliable signal.

"The process is slow, but judgement can't really be automated. It would be amazing if AI could do that for me."

— Recruiter

"Every resume looks the same now. AI wrote it to pass the filter, not to actually describe the person."

— Hiring manager

"Me and my hiring manager just don't agree on what good looks like. We waste so much time on that."

— Recruiter

"The process is slow, but judgement can't really be automated. It would be amazing if AI could do that for me."

— Recruiter

"Every resume looks the same now. AI wrote it to pass the filter, not to actually describe the person."

— Hiring manager

"Me and my hiring manager just don't agree on what good looks like. We waste so much time on that."

— Recruiter

5 interviews, 4 workshops.

5 interviews, 4 workshops.

How might we improve the efficiency and accuracy of candidate sourcing, especially as AI makes resume quality an unreliable signal?

Market Opportunity

Market Opportunity

A market of users begging for improvement.

A market of users begging for improvement.

0%
Recruiter time on manual screening
0%
Recruiters using keyword-only search
0%
Hires fail to perform as predicted
$0B
Global ATS market by 2028

Solution concept - my strategic direction

Solution concept - my strategic direction

So I reverse-engineered the perfect hire,
designed to never let you invest in the wrong person.

So I reverse-engineered the perfect hire,
designed to never let you invest in the wrong person.

01
Semantic search surfaces candidates
02
Recruiter reviews and hires
03
Hire performance feeds back
04
Future searches refine further

Top hires are identified through integrated post-hire performance data, which determines what 'best fit' looks like for each role. The more you hire and review, the better the algorithm predicts your next hire.

Top hires are identified through integrated post-hire performance data, which determines what 'best fit' looks like for each role. The more you hire and review, the better the algorithm predicts your next hire.

Semantic search, sharpened by every hire.

Semantic search, sharpened by every hire.

Post-Hire Performance Data

Lumeir is learning from your hires' performance in Workday to sharpen future candidate matching.

W
WorkdayHRIS / HCM

Sync performance reviews, ratings, and retention data.

Synced just now

Semantic search understands context within a candidate's profile beyond keyword matching. It also learns from outcomes. Every successful hire teaches the algorithm what a good match looks like, improving search results over time.

Semantic search understands context within a candidate's profile beyond keyword matching. It also learns from outcomes. Every successful hire teaches the algorithm what a good match looks like, improving search results over time.

Key features - team build

Key features - team build

Five key features. Each grounded in what we heard.

Five key features. Each grounded in what we heard.

Feature 01Semantic Search
Feature 02Review Matches
Feature 03Contextual Filters
Feature 04Compare Candidates
Feature 05Profile Annotation

Type in natural language or click predictive phrases to build your search. No boolean strings, no missed talent.

"Many recruiters are keyword search recruiters - they lack deep knowledge in the field."

Competitive positioning

Now, Lumeir wins where existing tools have no answer.

Competitive positioning

ToolSemantic searchEvidence scoresTeam alignmentAI transparency
LinkedIn
Greenhouse
Workday
Lever
Lumeir
15%
projected time-to-fill
5-10 min
saved per candidate review
20-25
candidates reviewed per day
20-50%
increase in trust toward AI

No existing ATS combines all five dimensions hiring teams need. This is Lumeir's advantage, and the design strategy was built around it.

No existing ATS combines all five dimensions hiring teams need. This is Lumeir's advantage, and the design strategy was built around it.

No existing ATS combines all five dimensions hiring teams need. This is Lumeir's advantage, and the design strategy was built around it.

"It just makes more sense to search in natural language. I'm honestly surprised LinkedIn hasn't already done this. Saves me from overthinking boolean string formats."

Recruiter

On semantic search

Hiring managers send requisition files. Recruiters manually re-read them every time they screen a candidate. I built an uploader that pulls requirements automatically and surfaces an AI analysis.

SoloFigma MakeFounder approved

Discovery & Research

2.5 months · Team

Define & Vibecode

2 weeks · Solo

Refine Concept

2 weeks · Team

Lead Team Build

3 months · I Led

Resume signal reliability

AI adoption risesReliableUnreliable

How might we improve the efficiency and accuracy of candidate sourcing, especially as AI makes resume quality an unreliable signal?

0%
Recruiter time on manual screening
0%
Recruiters using keyword-only search
0%
Hires fail to perform as predicted
$0B
Global ATS market by 2028
01
Semantic search surfaces candidates
02
Recruiter reviews and hires
03
Hire performance feeds back
04
Future searches refine further
Post-Hire Performance

Lumeir learns from hire performance in Workday to sharpen matching.

W
WorkdayHRIS

Syncs reviews, ratings, and retention data.

Synced just now
Feature 01Semantic Search
Feature 02Review Matches
Feature 03Contextual Filters
Feature 04Compare Candidates
Feature 05Profile Annotation

Type in natural language or click predictive phrases to build your search. No boolean strings, no missed talent.

"Many recruiters are keyword search recruiters - they lack deep knowledge in the field."

Competitive positioning

Semantic
Evidence
Team
AI
LinkedIn
Greenhouse
Workday
Lever
Lumeir

"It just makes more sense to search in natural language. I'm honestly surprised LinkedIn hasn't already done this. Saves me from overthinking boolean string formats."

Recruiter

On semantic search

Process — 6 months

We discovered together.
I decided where to go next.

Research was a team effort. Strategic decisions about where to go next was mine.

"

Amanda's prototype provided us a preview on the product direction, giving us the confidence to let the team move forward.

Tyler Bartholomew, CEO & Cofounder

"The process is slow, but judgement can't really be automated. It would be amazing if AI could do that for me."

— Recruiter

"Every resume looks the same now. AI wrote it to pass the filter, not to actually describe the person."

— Hiring manager

"Me and my hiring manager just don't agree on what good looks like. We waste so much time on that."

— Recruiter

5 interviews, 4 workshops.

Hiring managers send requisition files. Recruiters manually re-read them every time they screen a candidate. I built an uploader that pulls requirements from the file automatically, filters candidates against it, and surfaces an AI analysis - so the decision is ready before the recruiter even opens a profile.

SoloFigma MakeUser testedFounder approved

Once resumes could be optimized in seconds, the signal recruiters relied on for years collapsed almost overnight.

The problem

Existing process: Not efficient. Not accurate. Not aligned.

Process — 6 months

We discovered together.
I decided where to go next.

Research was a team effort. Strategic decisions about where to go next was mine.

15%
projected time-to-fill
5-10 min
saved per candidate review
20-25
candidates reviewed per day
20-50%
increase in trust toward AI

Impact and outcomes

Impact and outcomes

Validated, approved, and greenlit by the founder.

Validated, approved, and greenlit by the founder.

"

Amanda's prototype provided us a preview on the product direction, giving us the confidence to let the team move forward.

Tyler Bartholomew, CEO & Cofounder

15%
projected time-to-fill
5-10 min
saved per candidate review
20-25
candidates reviewed per day
20-50%
increase in trust toward AI