Loading the page…
For hiring teams
Screen every applicant. Read only the evidence.
Publish a role and SynergicHire reads each resume against it, screens the applicants you pick with an AI that follows your plan, and hands you assessments where every score cites what the candidate said. You make every call.
Senior backend
148 applied · 12 strong✓ 12 invites sent
Replies come to you
1Applicants
A ranking you can check, line by line.
Switch on the apply link and applicants land under Applied. Each resume is read against the role's competencies. Any verdict whose quote isn't in the resume is dropped, and the fit band is computed from what's left.
- Hidden instructions get flagged. White text telling a model to rate a resume highly is shown to you, not obeyed.
- Act on a whole band at once. Invite, hold or decline it as one recorded decision. A decline needs a reason and a sample opened first.
- Email is handled. Applicants get a receipt, invites come from SynergicHire with replies to you, and your decline reason is never in the email.
Senior Backend Engineer, Payments
148 applied · 139 read · 9 need review
- Amara Okaforstrong8 yrs
Led the move of order intake to an event queue; p99 from 4 s to 90 ms
- Diego Silvastrong7 yrs
Owned Postgres partitioning for 2B rows of payment events
- Hana Kimstrong6 yrs
Designed idempotent retry middleware used by 30 services
- Kenji Imaipartial5 yrs
Built internal Go services; no queueing work quoted
- Lena Fischerpartial4 yrs
Payments integrations at a marketplace
- Tom Beckerweak2 yrs
No quote found for distributed systems
- Unnamed.pdfweakInstructions found in resume
Hidden text asked the reader to rate this resume highly
Flagged and unread resumes are never included.
2Screening plans
Your hiring manager writes the plan once. The screener sticks to it.
A draft comes from the job description. The hiring manager edits it and approves it under their name. That version is frozen, so candidates screened a month apart were asked the same things.
Job description
Senior Backend Engineer, Payments
You'll own order intake end to end — queues, idempotent retries, and the Postgres underneath. We ship with AI agents daily and expect you to check what they write.
Screening plan v1Draft
3Reading a screen
Twelve minutes of reading instead of a 45-minute call.
The transcript sits against the plan, section by section, showing whether each one was covered, ran out of probes, or ran out of time. Drawings and code sit next to the answer they were made at. Then comes the assessment: pick a score to see what it rests on.
Fundamentals3.3 / 4
AI fluencygrade B
System reasoning · 4 Strong
Named the bottleneck, the fix and its failure mode unprompted.
cites 2 answers
Screener
Walk me through the order service you rebuilt. What broke first under load?Candidate
Intake. Synchronous writes to Postgres serialised on one hot row, so p99 went from 80 ms to four seconds at peak.Screener
Sketch what you moved to. What happens when that queue backs up?Candidate
We shed load at the edge first, then let intake lag — a late confirmation beats a dropped order. Idempotency keys make the retries safe. board · 6 shapes
Screener
How do you check code an agent writes for you on work like this?Candidate
I give it the failing test and the one file, read the diff before I accept it, and rerun the suite. If it touches the retry path I write the edge-case test myself.An A–E fluency grade
A letter from the AI-fluency core alone. It reads as a band, not as more precision than the evidence supports.
Advice, not a verdict
“Worth a technical round”, “Borderline — read it”, or “Not on this evidence”. You decide.
Voice, if you want it
Candidates can speak their answers. Audio is kept only if you turn it on, and every listen is audited.
4The pipeline
Every round, on one board.
A candidate here is a person, not a link you sent once. The board puts them under the round it's waiting on. The column you'll check most is the one for candidates whose rounds are all in and who are waiting on a decision.
Screening1
A. Weber
Senior frontend
Scheduled · Thu 10:00
Technical2
E. Moreau
Staff engineer
Awaiting feedback
J. Okafor
Senior backend
Invite sent
Hiring manager1
L. Berg
Staff engineer
Scheduled · Fri 15:30
HR1
S. Nakamura
Senior frontend
Scheduled · Mon 11:00
Additional1
T. Novak
Staff engineer
Panel · Wed 09:00
Decision1
R. Diallo
Staff engineer
4 of 4 written up
Closed1
M. Costa
Senior backend
Hired · signed by sarah@
Longest untouched, first
Cards sort by how long nothing has moved, so a candidate you forgot rises to the top instead of sinking under new ones.
Waiting on us
One filter for rounds that happened but were never written up, plus candidates whose rounds are all in and undecided.
Interviewers are told
They're emailed when assigned, moved or cancelled, and when a write-up is owed — never about a change they made themselves.
Feedback that compares
Every round is scored on the same things, every time.
Each kind of round has its own criteria: motivation and role fit on a screening call, ownership with the hiring manager, problem solving and AI fluency in the coding round. A write-up that skips one is refused, because half a scorecard makes two candidates impossible to compare.
A round is only finished when it is written up. Nobody can simply mark it done.
Technical round · write-up
Problem solving
AI fluency
Code quality
Communication
Overall4 / 5 · Yes
✓ Signed off — advance to the hiring manager round, by sarah@yourcompany.com
5The coding round
Five choices. One invite.
Pick a stack and a question, set the length, choose guarded or unguarded AI, and decide what is logged outside the code. Press Send invite and it goes out from SynergicHire.
The link only works for the candidate's own email address. If it's forwarded, consent is refused, so a leaked link can never become a session filed under the wrong person.
Technical round · set up
Stack
Node.js service
Duration
60 minutes
AI mode
Unguarded
Proctoring
Standard · focus and paste logged
Question
Add per-key rate limiting to the orders API. The failing tests describe the behaviour.to amara@hey.com · replies to youSend invite
After the round
Review the work, not just the result.
The report holds the whole session: code, whiteboard, AI transcript, agent edits, the app running, and how they used Plan, Build and Review. Code accepted from the AI is tinted, and the scaffolded project is left out because nobody wrote it.
1// src/routes/orders.ts
2import { rateLimit } from "../rateLimit";
3
4const allow = rateLimit(20, 60_000);
5
6export async function POST(req: Request) {
7 const key = req.headers.get("x-api-key") ?? "anon";
8 if (!allow(key)) {
9 return new Response("Too many requests", { status: 429 });
10 }
11 const order = parseOrder(await req.json()); // was: req.json() as Order
12 return Response.json(await createOrder(order), { status: 201 });
13}
typed · 9 linesaccepted from AI · 4 linesedited inside an AI region · 1 line
AI transcript full session
Candidate · 3:12
@orders.tsGuard the POST handler with the limiter keyed on x-api-key. Return 429 when it refuses.Assistant · 3:15
Added a key lookup with an anonymous fallback and an early 429 return.▣ accepted · 4 lines into orders.ts
Fluency signals
Prompt quality0.82Time to first prompt42sAI-generated share52%Verification0.90Over-reliance flags0Recovery1.00
Rubric rationale required
Problem decompositionRead the tests first, split limiter from route.
Prompt qualityTagged both files, named the expected shape.
Verification & skepticismRan the suite straight after accepting; caught the lockout bug.
Code qualityClean, but the pruned list is not saved on reject.
CommunicationExplained the fix; did not flag the trade-off unprompted.
Sign-off
Recommendation: lean yes
Signed by a named interviewer · recorded in the audit log
Assessed on two cores
Prompt precision, steering, inspection and verification — observed in the record, each score citing a step you can open.
Notes beside the evidence
Private to the hiring team and kept next to the work they describe.
Their IDE, full screen
Open the round's code in the IDE the candidate used, and run their app in its own tab.
1// src/routes/orders.ts
2import { rateLimit } from "../rateLimit";
3
4const allow = rateLimit(20, 60_000);
5
6export async function POST(req: Request) {
7 const key = req.headers.get("x-api-key") ?? "anon";
8 if (!allow(key)) {
9 return new Response("Too many requests", { status: 429 });
10 }
11 const order = parseOrder(await req.json()); // was: req.json() as Order
12 return Response.json(await createOrder(order), { status: 201 });
13}
typed · 9 linesaccepted from AI · 4 linesedited inside an AI region · 1 line
Authorship, visible
See what they typed and what they took.
Accepted completions, agent edits and code from chat are matched against the lines that survived into the final solution. A line the candidate rewrote after accepting it counts as theirs, because editing AI output until it's right is authorship too.
Your identity provider, three roles
SSO or a magic link through Supabase Auth — no local passwords. Recruiters own the pipeline, interviewers write up and sign off, admins do both plus exports.
Compliance exports are buttons
The append-only audit log as CSV, selection rates by role, seniority and source, and one row per decision for an independent auditor.
Integrity without surveillance
Focus changes and external pastes are shown to a person, never auto-flagged. Set minimal proctoring and neither is recorded.
Retention you set
Expired recordings are erased while scores and sign-offs stay. Declined applications expire too.
Per interview. No seats.
You pay for AI screens and coding rounds that run. Candidates, scheduling, other rounds and write-ups cost nothing. Interviewers, recruiters and admins are free.
See a whole hire in an afternoon.
Open a role, apply to it yourself from another tab, take the screen, and read the assessment where you started.