eval.

Product & roadmap

eval. today.
And what’s next.

Product overview / Roadmap: December 2026 — April 2027

From setting up a round to reviewing a candidate’s work, see how eval. fits together today — and how the next five months could build on it.

Start with the product

The product, in five slides.

In the current build

Set the round.
Keep control.

Shape the interview around the role you’re hiring for.

  • Role, skills & question mix
  • AI-generated questions or cloned rounds
  • Invites, timing & interview rules
Interview setupYour round, your rules

Backend Engineer

SkillsAPIs · Databases
DifficultyMedium + hard
ProgrammingTheoreticalMCQ
Moderator assignedDuration & expiryFullscreen
1 / 5Illustrative views & example data

Next: useful at every step of hiring.

Proposed target windows
Final scope and timing follow validation.

December

2026

01 / Consistency

Interview kits & evidence briefs

Planned

For companies

One approved plan.
A clearer reason for every rating.

Reuse role-specific interview plans and give reviewers a concise brief linked to the candidate’s actual work.

Delivery scope

  • Versioned kits with questions, timing, competencies, and scoring anchors.
  • A printable, single-interview brief with evidence links and reviewer approval.
  • A small set of company-reviewed answers to seed scoring calibration.

Success looks likeA team reuses a published kit, and a reviewer can trace a brief’s claims to saved evidence.

January

2027

02 / Operations

Recruiter workflows & candidate support

Planned

For companies

Less coordination.
More time for the right conversations.

Organize interview batches, follow up on outstanding work, and help candidates clarify the task within agreed boundaries.

Delivery scope

  • Explicit job/application links; CSV preview, batch invitations, and assignments.
  • Per-row failures, manual reminders, and retries with clear operational status.
  • Question-linked messaging and an opt-in AI clarification pilot that excludes solution guidance.

Success looks likeA recruiter can identify and retry failed rows without recreating successful invitations.

February

2027

03 / Real work

Work simulations & assessment calibration

Planned

For companies

See how candidates investigate,
fix, and verify.

Assess realistic debugging work across files. Check how grading changes against answers your reviewers have already assessed.

Delivery scope

  • Multi-file question generation, file navigation, and complete file-map submissions.
  • One selected runtime with isolated execution and public test feedback.
  • Reference answers, current/proposed grading comparisons, and human review of differences.

Success looks likeOne bounded debugging exercise runs end to end, with execution evidence and reviewable grading changes.

March

2027

04 / Connection

ATS screening workflows & team handoffs

Planned · ATS pilot

For companies

Screen candidates inside
your existing hiring flow.

An event in the organisation’s ATS triggers an Eval screening interview. Results return to the same candidate profile for the hiring team to review.

Delivery scope

  • Map the ATS job and stage to a screening plan; preserve the application ID.
  • Create and invite through a webhook to Eval’s API; retry without duplicate interviews.
  • Return reviewed results to the ATS; support candidate comparisons and next-round briefs.
  1. Your ATSCandidate reaches screening stage
  2. Webhook to EvalCandidate profile, job & stage
  3. Screening interviewCreate, invite & assess
  4. Back to the ATSScores, summary & report link

Success looks likeAn ATS webhook starts one screening round; its reviewed result returns to the correct candidate/application, with delivery and retries tracked.

April

2027

05 / Conversation

Conversational AI interviews

Proposed pilot

For companies

A natural conversation.
A structured assessment.

Let an AI interviewer ask, listen, and follow up within an approved plan, with a familiar, authorized voice and a report people can review.

Delivery scope

  • One spoken interview format: approved questions, bounded follow-ups, and interruption handling.
  • Correctable transcripts, confirmed answers, question-linked grading, and human review.
  • An authorized custom-voice trial, clear AI introduction, and per-interview voice/LLM usage tracking.

Success looks likeA complete spoken round produces saved, question-linked answers and a reviewable assessment.

One platform. More kinds of interviews.

Exploration / not yet scheduled

More roles. More realistic work.

Expand from the first debugging and spoken rounds into sales, support, product, and design work samples.

A broader hiring stack.

Extend the first ATS pilot into additional connectors and scheduling workflows, guided by company demand.

Richer conversational rounds.

Explore voice alongside coding and whiteboards, more languages, and additional authorized interview voices.