Skip to main content

PortfolioCase study

SR-01Live · team build

The Shravan's Round

He listens. Then he asks the follow-up.

A 20-minute AI mock PM interview fronted by Shravan's voice: real Indian product cases, cross-questions that follow your answer, and a free score card at the end. No login, no card.

The Shravan's Round landing page: “Rehearse the round before the round.” over a Start a free mock button and an amber waveform marked Listening.
  • 20 min

    one round, then a free score card

    Structural

  • 39/45

    the winning score in my avatar trial

    Measuredas of 10 Sep 2026

    Three vendors scored on lip latency, interruption and identity; the runner-up scored 35/45.

Evidence keyMeasuredStructural

Why it exists

A voice next to a photo reads as a chatbot, not an interviewer.

Candidates prepare for PM rounds with US textbook cases and no one to push back. The promise here is a round on the Indian cases Shravan teaches with, and it only works if it feels like a person is listening.

…a static photo next to a voice reads as a chatbot, not a person.

— Avatar layer Discovery PRD §1

The round

Set up, sit the round, read the score card.

A short setup with no login; then a 20-minute interview covering your story, one product case and metrics; then a score card with evidence quotes and one thing to fix, downloadable as a PDF.

The Shravan's Round landing page: “Rehearse the round before the round.” over a Start a free mock button and an amber waveform marked Listening.
The site's example score card: a “Round complete” summary with strongest area Product sense, 18 minutes, 8 questions and 5 follow-ups, beside a Download detailed report button.
The example score card (a made-up round)
The Shravan's Round on a phone: the headline, Start a free mock and See a sample score card buttons, and the listening waveform.
Phone: the same round

My part

Pick the face by measuring it, then let the waveform carry the brand.

  • Could have
    Pick an avatar vendor from demos
    Chose
    A scored trial of three vendors on one clip
    Because
    Lip latency, interruption and identity were measured the same way for each, and the best scorer went to the pilot.
  • Could have
    Build our own 3D face
    Chose
    A commercial photo-driven avatar behind our own interface
    Because
    Only single-view studio photos existed, so capture-based rendering was ruled out and the vendor stays swappable.
  • Could have
    A static logo
    Chose
    An animated listening ring
    Because
    The hero's waveform, bent into a ring, says “he is listening” in the nav and the favicon.

How it works

Your answer decides the next question.

  1. You answervoice in the browser, or text
  2. Classify the answeran LLM call
  3. Retrievehis teaching transcripts, pgvector
  4. Interviewer's next linethe follow-up
  5. Score carda second model writes the report
One answer through The Shravan's Round

Evidence

Live and free — no numbers yet.

The round is live and free to sit. The repository is private to the team. No candidates, sessions or scores have been recorded for this page.

What isn’t measured

  • No users or usage data are recorded.
  • A photo-driven avatar shows head and shoulders only, with no hand gestures.

Sources

Everything here is backed by a real artifact.

  • Readme
  • Research
  • Design
  • Live data

Evidence behind the The Shravan's Round case study

Private working documents are listed with what they support; only public sources link out.

  1. README

    Readme

    Supports: The flow, the stack and which layer is mine

  2. Avatar layer Discovery PRD and vendor trial

    Research11 Sep 2026

    Supports: The avatar problem, the three-vendor trial and its scores

  3. Brand mark notes

    Design15 Sep 2026

    Supports: The animated listening-ring logo and favicon

  4. Live captures

    Live data17 Oct 2026

    Supports: The landing and example score card shown here

Links and next steps

Team build · avatar, brand mark and filmsSep–Oct 2026Live · team build