← Back to hackathon details

Problem statements

Pick one of the five problem statements below. Every one of them is built and used with your voice.

Submit your projectCloses 31 Oct, 11:59 PM IST

Applies to every problem statement

Build with your voice

Every team builds its solution using Wispr Flow. How well you use it counts for 20% of your score.

Download Wispr Flow

  1. Build it by voice

    Use Wispr Flow to dictate your prompts, specs and fixes into the AI builder of your choice, such as Lovable, Bolt, Cursor, Replit or Claude, instead of typing them.

  2. Make it work with voice

    The main input to your product is natural speech captured through Wispr Flow. Real speech has filler words, digressions and self-corrections, so your agent has to turn it into clear, structured output.

  1. 01Media: content repurposing studio with brand memory

    Teams must build:

    • long video or podcast ingestion;
    • clip selection with reasons given for each choice;
    • captions and copy tailored to each platform;
    • a brand-voice memory that learns from the creator’s edits;
    • a content calendar;
    • a performance feedback loop that changes future suggestions based on what actually performed.
  2. 02Adaptive exam-prep tutor with a learner model

    Generic AI tutors answer questions but don’t track what a student actually understands.

    Teams must build:

    • a diagnostic test;
    • a concept graph for one syllabus, such as Class 10 physics;
    • a learner model that updates mastery for each concept after every answer;
    • question generation targeted at weak concepts, with difficulty calibration;
    • misconception detection from the reasoning in wrong answers, not just from the wrong final answer;
    • spaced-repetition scheduling;
    • a teacher dashboard showing class-wide gaps.
  3. 03Meeting-to-execution agent

    Decisions made in meetings get lost.

    Teams must build:

    • transcript ingestion with speaker attribution;
    • extraction of decisions, action items, owners, and deadlines;
    • detection of conflicts with earlier decisions stored in memory;
    • creation of tasks in a tracker through an integration;
    • reminders before deadlines;
    • a weekly digest per person;
    • a confirmation step before anything gets posted.

    The cross-meeting memory and the conflict detection are what make this a real system rather than a transcript summarizer.

  4. 04Personal finance coach with a living money model

    Most budgeting apps categorize transactions and stop there. They don’t understand a person’s goals or adapt as life changes.

    Teams must build:

    • ingestion of bank statements in messy formats (PDF, CSV, and transaction SMS), with automatic categorization;
    • a correction loop where the user fixes a category and the system learns the rule for future transactions;
    • detection of recurring payments and subscriptions, with alerts when one changes or goes unused;
    • a persistent financial profile covering income, fixed costs, emergency fund, and goals with target dates;
    • a goal planner with what-if simulation, such as “what if I cut dining out by 30%” or “what if my rent goes up”;
    • nudges when spending drifts off plan;
    • a monthly review report;
    • clear guardrails, so the coach educates and models scenarios but doesn’t recommend specific securities.
  5. 05Career pivot coach

    Teams must build:

    • resume and target-role intake;
    • a skill-gap map built from real job descriptions;
    • a week-by-week learning plan;
    • portfolio project suggestions matched to the gaps;
    • mock interviews scored against a rubric;
    • progress tracking that adjusts the plan;
    • an application tracker that learns which kinds of roles get responses.