TestMe platform · Onboarding early teams

Technical interviews built for how engineers actually work now.

TestMe is Genspur's AI-native interview platform. Candidates read real code, direct an AI coding agent, review generated changes, and design systems while a live AI interviewer asks follow-up questions. Your team gets an evidence-based report for every round.

Talk to us See the product

Code review

src/invites/accept.ts

18−return invite.user

18+if (invite.expired()) throw new Gone()

19+audit.log({ action, actor })

20 return invite.user

The agent added an expiry check. Ready to merge?

Almost. Expiry uses server time, so I'd test clock skew first.

Spotted edge case · explained the risk

A quick look at four of the rounds. Pick one, or scroll down for real product screens.

Product tour

See a TestMe interview, round by round.

These screens come from a sample five-round interview for a fictional company, Acme Engineering. The AI interviewer, coding agent and evaluator responses are real model output.

Build an interview in minutes

Recruiters pick topics, choose which rounds to include and in what order, set time limits, proctoring and voice mode, and publish a private invite link.

TestMe session builder showing the five round types: MCQ, Code Reading, Agentic Prompting, Code Review and System Design, with topic chips for AI Engineering Basics and Distributed Systems.

Technical knowledge, framed as real decisions

Multiple-choice questions describe a production situation and ask for the most targeted fix, so they test judgment rather than memorised trivia.

A distributed systems question about a user seeing a stale profile from a read replica, with four answer options and the targeted fix selected.

Code reading with a live AI interviewer

The candidate explains unfamiliar code in a read-only editor. The AI interviewer listens, by text or voice, and asks follow-up questions in real time.

Code reading round: a Python model-routing helper in a read-only editor beside the AI interviewer panel, where the candidate explains that the fallback model skips the allowlist check.

Agentic prompting: direct an AI coding agent

Candidates get a feature request and a codebase, then tell a simulated coding agent what to change. The agent responds with a diff, and every prompt iteration is recorded.

Agentic prompting round: a side-by-side agent code diff above the candidate's prompt asking the agent to make job deduplication tenant-safe and add focused tests.

Code review of a generated change

Candidates inspect a diff with planted issues, select changed lines, rate severity and explain the impact, as they would on a real pull request.

Code review round: a read-only diff that replaces structured tool dispatch with text parsing, and a major-severity annotation explaining the prompt-injection risk.

System design on a shared whiteboard

Candidates sketch an architecture and talk it through. The AI interviewer reads the diagram and the explanation, then probes the tradeoffs.

System design round for a global chat system: a whiteboard diagram of clients, a WebSocket gateway, a message log, a presence cache and a replicated store, with the AI interviewer asking about ordering and latency tradeoffs.

What we build

Two focused products for technical teams.

Core product · TestMe

AI-native technical interview platform

Interview rounds built around real engineering work: technical decisions, reading code, directing AI coding agents, reviewing changes, and system design. Each round produces a clear, evidence-based record of what the candidate did and why.

Project based

Applied AI workflows

Focused tools for teams with recurring technical review or decision work, where AI can remove real friction. We scope each one with you around a specific problem.

How it works

Five rounds, each testing a different skill.

Instead of one generic coding puzzle, each round has its own context, tools, and rubric. Recruiters choose the rounds that fit the role.

  1. Technical knowledge

    Scenario-based questions on the topics that matter for the role.

    Shows judgment

  2. Code reading

    The candidate explains unfamiliar code to a live AI interviewer.

    Shows comprehension

  3. Agentic prompting

    They direct an AI coding agent and correct its output when it goes wrong.

    Shows direction

  4. Code review

    They inspect a generated change, find the risks, and explain the fix.

    Shows rigor

  5. System design

    They sketch an architecture on a whiteboard and defend the tradeoffs.

    Shows systems thinking

No code execution, by design. Code is always shown in read-only editors and nothing a candidate submits is run. Rounds test understanding, direction and review, the skills that matter when AI writes much of the code.

Evidence-based evaluation

Reports your interviewers can actually discuss.

Every question carries its own rubric. After the interview, TestMe shows which concepts the candidate covered, which issues they found, and the conversation behind each score. Hiring decisions stay with your team.

Code reading analysis listing five rubric concepts the candidate covered, including that the fallback model bypasses allowlist validation, with no missing concepts.
Rubric coverage for a code reading round.
Code review analysis listing three issues found with line numbers: two security issues and one authorization issue.
Issues found in a code review, mapped to lines.
System design discussion transcript showing the candidate's design explanation beside the AI interviewer's follow-up question.
The full design discussion, turn by turn.

Platform

Everything a modern technical interview needs.

  • Live AI interviewer

    A realtime interviewer joins each live round over WebRTC, asks follow-up questions and adapts to the candidate's answers.

  • Voice mode

    Candidates can talk instead of type. Speech-to-text, the interviewer's reply and text-to-speech run in one low-latency pipeline, with a full transcript.

  • Simulated coding agent

    A coding agent that responds to the candidate's prompts with real diffs, so prompting skill shows up in the result, not in a description of it.

  • Vision-assisted design review

    Whiteboard snapshots are read by a vision model, so the evaluation covers what the candidate drew as well as what they said.

  • Privacy-first proctoring

    Face-presence detection runs on the candidate's device, alongside tab and focus signals. No face recognition, emotion inference or keystroke biometrics.

  • Recruiter workspace

    Session builder, private invite links, live monitoring, and a results view with per-round evidence and integrity signals.

  • Practice mode

    Candidates can practise any round type on their own and get streaming feedback on what went well, what to improve, and one concrete tip.

  • Verifiable certificates

    Certification exams issue PDF certificates with a QR code that links to a public verification page.

  • Secure access

    Role-based access for admins, recruiters and candidates, invite-only recruiter accounts, and Google sign-in for candidates.

Engineering

Built on a production-grade cloud stack.

TestMe runs on Microsoft Azure with infrastructure as code, automated deployment pipelines, and managed secrets. Interview progress is durable, so a dropped connection never loses a candidate's work.

Application

  • Next.js
  • TypeScript
  • FastAPI
  • Python 3.12
  • Async SQLAlchemy
  • Pydantic v2
  • Monaco editor
  • Excalidraw

AI and realtime

  • Anthropic Claude API
  • Claude tool use
  • Claude vision
  • OpenAI GPT-4o
  • LangGraph
  • LiveKit WebRTC
  • Whisper STT
  • Neural TTS
  • MediaPipe (on-device)

Provider-agnostic model layer: every model call goes through one client, so each task (interviewer, coding agent, evaluator, whiteboard review) can run on Claude or GPT-4o.

Data and infrastructure

  • Azure Container Apps
  • PostgreSQL + pgvector
  • Redis
  • Azure Service Bus
  • Azure Blob Storage
  • Azure Key Vault
  • Bicep
  • GitHub Actions
  • Append-only interview record. Every answer, prompt, annotation and AI message is stored as an immutable event, so reports can always be traced back to what happened.
  • Durable workflow. Section progress lives in the database, so candidates can rejoin an interview and continue where they left off.
  • Asynchronous evaluation. Scoring runs in queued workers that scale independently of the interview itself.
  • Hardened by default. Rate limiting, health and readiness checks, structured request logging, and managed identities with secrets in Key Vault.

Milestones

From proof of concept to a full platform in six months.

We build in small, reviewed steps against a written roadmap, and every step is tested and documented as it lands.

  1. Proof of concept

    First working prototype of a technical interview built around reading code.

  2. Genspur LLP incorporated

    Registered with the Registrar of Companies, Bengaluru, on 1 April 2026.

  3. V1 · Core platform live on Azure

    Project scaffold, data model, authentication, organisations and recruiter dashboard, question bank, session builder, candidate invites, the interview room, automated grading, and Azure infrastructure with CI/CD.

  4. V2 · AI interview rounds

    Realtime rooms on LiveKit, code reading, agentic prompting and code review rounds, durable interview workflow, asynchronous evaluation, invite-only recruiter access, managed question releases, and per-environment realtime controls.

  5. V3 · Trust, design and experience

    Hardened admin provisioning, recruiter password reset, Google sign-in for candidates, a full UI redesign, blob storage, the system design whiteboard, on-device proctoring, and credible evaluation reports.

  6. V4 · Voice and practice

    Voice mode across live rounds, self-serve practice mode with streaming feedback, certification with QR-verifiable certificates, and production hardening.

  7. V5 · Question library expansion

    A clear-colleague writing standard for every question, a full rewrite of the original 2,168-question bank, and expansion to 7,500 authored questions across 41 topics, including Java and Spring, C# and .NET, Go, Rust, event streaming, and data platforms.

Company

About Genspur.

Genspur is an independent software company in Mysuru, India. We build AI products for technical teams, starting with how they hire engineers.

Legal name
Genspur LLP
LLPIN
ACW-8570
Incorporated
Registrar
Registrar of Companies, Bengaluru (Ministry of Corporate Affairs, India)
Registered office
Srirampura 2nd Stage, Mysuru, Karnataka 570023, India
Activity
Computer programming, consultancy and related activities

Designated Partner

Smitha S

Leads Genspur and is the first point of contact for teams interested in TestMe.

LinkedIn profile

Company details are public record. View the registry listing.

FAQ

Common questions.

What is TestMe?

TestMe is Genspur's technical interview platform. Candidates work through realistic rounds (technical knowledge, code reading, agentic prompting, code review, and system design) with a live AI interviewer, and hiring teams get a structured report of what the candidate did and why.

Do candidates have to write and run code?

No. Code is always shown in read-only editors and nothing is executed. Candidates explain code, direct an AI coding agent in plain language, review generated changes, and design systems, which is how much of engineering work happens today.

How are candidates evaluated?

Every round is scored against a rubric written for that question. Reports list which concepts the candidate covered, which issues they found, and the evidence behind each score, so interviewers discuss specifics rather than a single number. Hiring decisions stay with your team.

How does proctoring treat candidate privacy?

Face presence detection runs on the candidate's device in the browser, and only integrity signals are recorded. There is no face recognition, emotion inference, or keystroke biometrics.

How can my team get started?

Email [email protected]. We are onboarding early teams and will set up a workspace around your roles and topics.

Contact

Hiring engineers, or working on an AI workflow?

We'd be glad to show you TestMe and hear what your team is seeing. Email is the fastest way to reach us.

Email [email protected] LinkedIn Smitha S