v1.0 · Now in private beta

Hire the engineer,not the resume.

Hyrrd reads real work — commits, PRs, architecture decisions, career trajectory — and surfaces the engineers who will actually thrive in your role. Built for a world where resumes are cheap and signal is everything.

No coding tests No keyword filters No AI-resume noise
hyrrd · match engine v3
thinking

// signal extraction

Reading 47 repositories from github.com/maya-r

Analyzing 1,284 commits, 312 PRs, 86 code reviews

Building tech-stack depth vector across 5 domains

Matching against 12 active roles in your pipeline

MR

Maya R.

Senior Backend Engineer · 6y · Berlin

top 3%
Capability fit0
Growth fit0
Domain fit0
Culture fit0

Strong match. Deep PostgreSQL + distributed systems signal from kafka-connect contributions. Startup ownership pattern across 3 prior roles.

Trusted by engineering-first teams hiring across > 30 stacks

TypeScriptGoRustPythonKubernetesPostgreSQLReactNext.jsNode.jsTerraformAWSKafkaGraphQLRedisClickHouseSwiftKotlinWebAssemblyElixirSparkPyTorchtRPCPrismaOpenTelemetryTypeScriptGoRustPythonKubernetesPostgreSQLReactNext.jsNode.jsTerraformAWSKafkaGraphQLRedisClickHouseSwiftKotlinWebAssemblyElixirSparkPyTorchtRPCPrismaOpenTelemetry

The problem

Hiring engineers is broken — and AI made it worse.

In a world flooded with AI-generated resumes, the real advantage isn't resume writing. It's signal extraction.

01

Resume spam from AI

Every recruiter inbox is now flooded with AI-polished, keyword-stuffed resumes. Keyword filters get gamed in a single prompt.

02

Keyword filters miss talent

Self-taught engineers, career-switchers, and OSS contributors without big-name companies get filtered out before a human ever looks.

03

Senior GitHub is mostly private

The most experienced engineers have the least public code. Star counts and follower numbers tell you almost nothing about depth.

04

Screening is brutal labor

Recruiters scan 200+ resumes per role to find 5 worth a phone screen. That's not a process — that's a tax on hiring.

// how it works

A hiring engine that reads code, not keywords.

Four stages. One feedback loop. Continuously learning from every hire.

step 01

Ingest real work

GitHub repos, PRs, commits, code reviews, resume narrative, career trajectory — all ingested via OAuth or upload.

$ rh ingest --source github,resume --depth 90d
step 02

Extract signal

Claude-powered pipelines build a Candidate Intelligence Graph: code complexity, ownership, authenticity, domain depth.

→ candidate_signals: 47 dimensions, confidence 0.92
step 03

Match multidimensionally

Capability, growth, domain, culture — each scored independently. Embeddings + learned re-ranking, never keywords.

match.score = f(capability, growth, domain, culture)
step 04

Learn from outcomes

Every interview, offer, and 90-day retention rating feeds the loop. Signals that predicted real hires get more weight.

model.retrain ← outcome(hire_quality_90d)
Candidate Intelligence Graph

47 signals.
One honest score.

Most platforms have one or two signals — years of experience, tech tags — and pretend those are enough. We extract dozens of orthogonal signals from real work, weight them against real hiring outcomes, and never expose the scoring methodology to candidates.

// match_score.breakdown

Capability fit94
Growth fit88
Domain fit91
Culture fit82
Compositetop 3%

Code complexity

Per-domain depth scoring from architecture patterns and dependency choices.

Contribution authenticity

Detects fork-and-rename, bulk commits, template scaffolds. Real work only.

Ownership signals

Distinguishes "led / architected / owned" from "helped / assisted" — at scale.

Tech stack depth

Multi-year vector per stack, not a self-reported skill list with 14 logos.

Timeline consistency

Cross-references claimed tech with actual GitHub timeline. Catches inflation.

Career trajectory

Company-tier normalization + role progression rate. No more pedigree blindness.

Collaboration footprint

PR descriptions, review comments, response quality — communication is a signal.

Bias audit trail

Every automated ranking decision is logged. Demographic correlation alerts built in.

For recruiters

Talk to your pipeline like it's a teammate.

Natural-language search. AI-ranked shortlists. Per-candidate summaries written by a model that's read every line of their public code. Screening time drops by 60% — quality goes up.

  • Ranked candidate lists per role, refreshed every hour
  • AI summary cards: 3-sentence capability, strengths, gaps, interview focus
  • NL search: "Backend engineers with PG optimization & startup background"
  • Auto-shortlist + one-click "not a fit" — feedback trains the model
  • Pipeline funnel + signal-quality analytics built in
Open the recruiter dashboard
> backend engineers with PostgreSQL optimization, startup background, 2+ years
SP

Sasha P.

Backend · 4y · ex-seed startup

top 2%
96
DK

Daniel K.

Platform · 3y · scale-up

top 5%
91
AN

Ari N.

Backend · 2y · YC W24

top 8%
87
LJ

Lin J.

Backend · 5y · solo founder

top 11%
85
248 more candidates ranked · refreshed 4 min ago→ shortlist top 20
78%

Profile strength

+12% since linking GitHub

GitHub connected · 24 repos analyzed

Resume parsed · 3 prior roles

Add 1 more project to unlock 4 new role types

Link LeetCode / Codeforces (optional, +signal)

top picks for you

3 new roles match your distributed systems + PostgreSQL depth. Series A, remote OK.

For candidates

Your code is your resume.

Spend zero minutes writing bullet points. Connect GitHub, drop in a resume, and we'll do the rest — surfacing roles where your actual depth matches what the team actually needs. No AI-resume generation. No score-gaming. Just signal that's already in your work.

  • Opaque profile-strength prompts — gamified, never gameable
  • AI-curated weekly top picks with plain-English match reasoning
  • Smart apply assist: surfaces which projects to highlight per role
  • Built-in scheduling with calendar sync + ICS reminders
  • Privacy-first: opt out of any signal, export or delete anytime
Sign in with GitHub

60%

Screening time saved

Recruiters spend the hours on real conversations.

Shortlist-to-interview rate

Industry baseline: 10%. Ours: 30%+.

47

Signals per candidate

Code, career, behavioral, outcome — orthogonal by design.

0

Keyword filters

No "5 years of React" gating. Real depth, real comparison.

// faq

Questions every hiring team asks.

How is this different from a traditional ATS?+

ATS systems match resume keywords. Hyrrd reads the underlying work — public repos, commit patterns, PR descriptions, architecture choices — and produces a multi-dimensional match score across capability, growth fit, domain fit, and culture fit.

Can candidates game the match score?+

Match scores are derived from dozens of orthogonal signals — code complexity, ownership verbs, contribution authenticity, timeline consistency — and the methodology is hidden from candidates. No single metric is gameable, and we monitor for manipulation patterns.

What about engineers whose best work is private?+

Senior engineers' public GitHub is often light. We blend GitHub signal with resume depth, impact statements, ownership signals, and company-tier normalization. The model weights whichever signal is more reliable for each candidate.

Do candidates see their score?+

No. Candidates see profile-strength prompts ("add 2 more projects to unlock more matches") but never the raw score. This protects signal authenticity and stops the platform from devolving into score-optimization games.

How do you prevent bias?+

Every automated ranking decision is logged. We monitor demographic correlation with output rankings and alert when specific signal types correlate with protected characteristics. Audit logs are exportable for compliance review.

Is this only for senior engineers?+

No. Match weights adapt by seniority. For freshers, GitHub activity, side projects, and competitive coding profiles dominate. For staff+ engineers, architecture ownership and impact statements dominate. The signal mix is learned, not hardcoded.

// the hiring intelligence layer

Stop reading resumes.
Start reading signal.

Join the private beta. Two-minute setup. First ranked candidates within 10 minutes of posting your first role.