3.0 Screen

Every application read
and scored

Progressive scoring.
Not one-shot filtering.

CV review, email Q&A, written tests, and AI video pre-screen — each layer lifts confidence. Every score is numbered against the Brief's rubric with reasoning.

FIG. 1.1
Progressive scoreAlex Chen
CV review
Low45
Email Q&A
Low55
Written test
Medium75
Video pre-screen
High88
Composite88 / 100

Score updates as data arrives

Every candidate starts at baseline and progresses through CV review, Q&A, tests, and video. Confidence tracks with data volume — low, medium, or high — surfaced explicitly.

FIG. 1.2
Over-qualified

“Staff-level at FAANG — may expect higher scope than IC4 band”

Salary mismatch

“Expected $240k vs. budget ceiling $195k — 23% over”

Employment gap

“18-month gap (2021–2022), no explanation provided in CV”

Visa sponsorship

“Current H-1B expires Aug 2026 — transfer required within 60 days”

Red flags with cited evidence

Over-qualification, anomalies, and near-misses surfaced with evidence. Nothing is silently auto-rejected — you decide on every edge case.

FIG. 1.3
84High confidence
CVEmail Q&AWritten testVideo
71Medium confidence
CVEmail Q&AWritten testVideo
63Low confidence
CVEmail Q&AWritten testVideo

Three-tier confidence

Confidence reflects data volume: low, medium, or high. You always know how solid a score is based on what data is available.

FIG. 1.4
Screen
Tabs
Paste
Webcam
Time
AI

Test integrity detection

Monitors screen-leave events, tab switching, webcam, copy-paste, automation, and time anomalies. AI cheating detection in video pre-screens.

FIG. 1.5
Blind mode
ON
redacted
redacted
redacted

Skills & scores visible

Technical
92
Communication
85
Cognitive
88
Problem solving
79
System design
91

Optional blind screening

Hide name, photo, university, and location if you want it. Employer-configurable per role. Reveals only after the shortlist is approved.

Not keyword-matching.
Actually reading.

Resume, cover letter, links, previous work. Eligibility pre-screening before rubric scoring. Custom tests from plain-English descriptions.

FIG. 1.6
ApplicationFull read complete

Resume

Parsed · Not keyword-matched

Eligibility pre-screen

Right to Work (UK)
Salary expectation
Availability (2 weeks)
Location (London)

Links reviewed

GitHub (47 repos)
Blog (12 posts)
LinkedIn

Full application, fully read

Resume, cover letter, links, previous work — not keyword-matching, not just parsing. Eligibility pre-screening runs before rubric scoring so unqualified candidates don't waste pipeline time.

FIG. 1.7
test-builder
>

Test event pipeline handling with exactly-once delivery

Generated 5 scenarios · rubric attached

01Scenario: 10M msgs/day throughput test
02Scenario: Exactly-once under partition failure
03Scenario: Consumer group rebalancing
04Scenario: Dead-letter queue overflow handling
05Scenario: Cross-region failover with state sync

Custom test builder

Tell Rob what to test in plain English. He generates scenarios and rubrics. Built-in library covers English, negotiation, support, and technical roles.

FIG. 1.8
1
You override
Maya PatelScore: 68
RejectAdvance
2
Rob learns

+ Weight “migration experience” higher

+ Reduce penalty for non-traditional path

3
Scoring adaptsNext batch
Similar candidates:6881

Override learning loop

Every override you make feeds back into the rubric. Rob's scoring adapts to your pattern over time — getting sharper with every decision.

Surfaces what matters.
You decide the rest.

Rob never auto-rejects. Edge cases are flagged for human judgement. Audit trail on every decision so you can defend the call.

FIG. 1.9
Screening pipeline

200 applications

All read overnight

Sorted, reasoned pile

In your inbox by morning

You decide edge cases

Nothing silently binned

A sorted pile, not a screening marathon

200 applications read overnight. A reasoned, sorted pile in your inbox by morning. Edge cases flagged for your judgement — every decision has an audit trail.

FIG. 1.10
Tech 92
Comm 87
0:00Q1: Walk me through a migration you led
2:34Q2: How do you handle delivery guarantees?
5:12Q3: Describe your ideal team structure

AI video pre-screen

3–9 questions with per-response evaluation across six dimensions. Candidates can redo any response before submitting.

FIG. 1.11
audit.log
[09:14:02]

ADVANCED — Score 88, high confidence (by You)

[09:14:18]

REJECTED — Salary 2x budget, scope mismatch (by You)

[09:14:33]

FLAGGED — Near-miss, 1 must-have gap (by Rob)

[09:14:41]

ADVANCED — Score 91, override accepted (by You)

[09:15:02]

REJECTED — Bot detected, 0 integrity score (by Rob)

[09:15:14]

FLAGGED — Employment gap, no explanation (by Rob)

Audit trail on every decision

Every decision is logged with reasoning. “Why was this candidate rejected?” always has a rubric-cited answer you can defend.

What Screen
replaces

No more screening 200 resumes at 11pm, random rejections, or good candidates lost in the pile.

FIG. 1.12
Today

11:47 PM · 200 unread

“Rejected — doesn't have Kafka on resume”

“Good candidate lost in the pile”

Audit trailNone

From 11pm marathons to morning clarity

Late-night screening, keyword-filtering, and random rejections are replaced by a progressive scoring model that reads every application and gives you a defensible, sorted pile by morning.

FIG. 1.13

No auto-reject

Every candidate gets a human decision

No demographics

Name, age, gender, ethnicity never scored

No hidden dims

All scoring criteria visible and auditable

Flags, not rejects

Edge cases flagged — never silently dropped

No auto-reject, no bias

No demographic factors. No hidden dimensions. No silent binning. The employer is always the safety net — you decide on every edge case.

FIG. 1.14
Right to Work
PASS
Salary range
PASS
Availability
PASS
Location
BLOCKED

Eligibility pre-screening

Right to Work, visa status, salary expectations, availability, and location preference — checked before rubric scoring runs so ineligible candidates don't waste pipeline time.

Under the hood

Everything Screen handles, from progressive scoring to test integrity.

Progressive AI scoring

CV → Q&A → tests → video, updating in real-time as data arrives.

Rubric-based scoring

Per-criterion reasoning tied to Brief must-haves and nice-to-haves.

Five scoring dimensions

Technical, communication, cognitive, experience, motivation & fit.

Three-tier confidence

Low, medium, high — surfaced in Rob's framing, not hidden.

Built-in test library

English, negotiation, customer support, and technical roles.

Custom test builder

Plain-English description → Rob generates scenarios and rubrics.

AI video pre-screen

3–9 questions with per-response evaluation across 6 dimensions.

Test integrity detection

Tab switch, webcam, copy-paste, automation, time anomalies.

AI cheating detection

Catches candidates generating answers live in video pre-screens.

Eligibility pre-screening

Right to Work, visa, salary, availability, location — before rubric.

Optional blind screening

Name, photo, university, location hidden. Configurable per role.

Anti-spam and bot detection

Filters inbound applications before they reach your queue.

Override learning loop

Rubrics adapt to your decisions. Rob gets sharper over time.

Red-flag surfacing

Cited evidence, nothing silently binned. You decide edge cases.

No auto-reject

No demographic factors, no hidden dimensions. Employer is the safety net.

Questions
and answers

Will candidates get auto-rejected?

No. Rob never auto-rejects. Edge cases, red flags, and over-qualification are surfaced with cited evidence — you decide on every one. The employer is the safety net.

What if the AI score is wrong?

Every score has reasoning attached. Confidence indicators tell you how solid the score is based on the data available. If you disagree, you override — Rob logs the direction and reason, and the rubric learns from your pattern over time.

Do you screen for things you shouldn’t?

No demographic factors. Ever. Screening logic is built to surface qualifications, not demographic proxies. Bias audits run against every rubric. Decisions are explainable — “why was this candidate rejected?” always has a rubric-cited answer.

Can candidates cheat the AI pre-screen?

AI usage detection catches candidates generating answers live. Test integrity detection monitors screen-leave events, tab switching, webcam, copy-paste, automation, and time anomalies. Candidates can redo any response before submitting, but Rob sees the redo pattern too.

What does blind screening hide?

Name, photo, university, and location — optional, per role. Employer-configurable. Reveals only after the shortlist is approved.

What about roles with specific hard requirements?

Eligibility pre-screening covers Right to Work, visa status, salary expectations, availability, and location preference before the rubric scoring runs. Candidates who can’t take the job don’t waste your pipeline time.

1.0 Brief

Turn a role into a hiring plan

2.0 Source

Find candidates who actually fit

4.0 Interview

Interviews that build on context

5.0 Reference

References that reveal the truth

6.0 Hire

Close the hire, not just the loop

“We stopped chasing candidates and started closing them. Rob handled the entire pipeline while we focused on building.”

Mira Johansson

VP Engineering, Vercel

“Six hires in two months with zero recruiter hours. That's not an optimization — it's a different model.”

Daniel Kraft

Head of Engineering, Ramp

Jobsly powers hiring for fast-growing teams. From early-stage startups to scaling enterprises.

Customer stories

Built for the future.
Available today.