Aarav Kalra, home
Q2 The Film Room ThinkLinks case study
← Back to the game

The Film Room · Case study

ThinkLinks

Teaches K–12 students how AI works. Every lesson is assembled on the spot for one student, based on how that student learns.

Role
Co-founder & CTO
Team
Me + Hemant (outreach, curriculum)
When
Grades 11–12
school pilots
students
cost per student

00:00 · The problem

We live in Silicon Valley. Google, OpenAI, and Nvidia are basically in our backyard. Yet as juniors, we had zero AI classes. Hemant and I have been friends since we were kids, and while messing around with AI tools we realized how confusing it was even for us. If we were struggling, how prepared were younger students?

01:12 · The insight

Ask how a kid learns, not what they know.

Asks how a student likes to learn — formats, pacing, language comfort, accessibility needs, interests — not what they already know.

Evolves with every action: replaying audio, skipping a video, getting stuck, leaving a lesson.

02:30 · The architecture

Three agents, one lesson

Input Learner profile Formats, pacing, language comfort, accessibility needs, interests. Updated by every action.
Agent 1Curriculum agentPicks the concept
Agent 2Teaching agentPicks the format: text, audio, visuals, simulations, games
Agent 3Assessment agentRuns low-pressure checks
Output The lesson Assembled live for one student, not pulled from a fixed library.
Feedback loop Their next move — replaying audio, skipping a video, getting stuck, leaving a lesson — feeds back into the profile.
ThinkLinks architecture: the learner profile feeds three agents, which assemble a lesson; the student’s actions feed back into the profile.
  1. 01OnboardingAsks how a student likes to learn — formats, pacing, language comfort, accessibility needs, interests — not what they already know.
  2. 02Learner profileEvolves with every action: replaying audio, skipping a video, getting stuck, leaving a lesson.
  3. 03Three agentsCurriculum agent picks the concept. Teaching agent picks the format — text, audio, visuals, simulations, games. Assessment agent runs low-pressure checks.
  4. 04The lessonAssembled live for one student, not pulled from a fixed library. Their next move feeds back into the profile.
ThinkLinks onboarding, step 1 of 4: “What do you love, Aarav?” with a grid of interests to pick from — sports, gaming, art, music, science, space, tech, coding and more.
Onboarding: pick what you love
Step 3 of 4: “How do you like to learn?” with choices for format (read, watch, quiz, play), pace (take it slow, just right, challenge me) and what helps it click (real examples, step by step, as a story).
Onboarding: how you like to learn
The accessibility panel: read this page aloud, a reading level switch (easier, just right, advanced), and toggles for read aloud, easy-read font, calm mode, bigger text, captions and high contrast.
Accessibility tools
The lesson “What is AI, really?” on its Play step, after Read and Watch and before Check: a side-scrolling game through History, Space and Science where a robot grabs things that really use AI and dodges what doesn’t.
A lesson: “What is AI, really?”
A “Let’s switch it up” prompt: “This isn’t quite landing — try Play instead?” with a Switch now button and a Keep Read option.
Switching formats mid-lesson

03:45 · Pilot results

Three schools, 200+ students

Piloted in 3 schools with 200+ elementary and middle schoolers, parents, and teachers — plus families in India, where AI is taught earlier.

04:20 · What changed after feedback

  • What we heard: early lessons still taught every student the same way.
  • We reworked onboarding, the adaptive logic, formats, and accessibility.
  • Teachers get a classroom dashboard of mastery and engagement.

05:10 · Business model

Free classroom trials convert to school and district licenses; parents subscribe at home.

  • Represented the USA at the SAGE Global Entrepreneurship World Cup in Vietnam — Top 6 in the world.
  • Mott Million Dollar Challenge semifinalist (top 60 of 3,700+, $7,750).
  • 2nd at Ferris State’s Bulldog Pitch.

05:55 · The patent

First named inventor

U.S. provisional patent: disclosure-free adaptive accessibility.

  • The full-stack product, from the ground up.
  • The multi-agent system and its infrastructure.
  • Technology, product direction, and business strategy as CTO.

06:30 · Press

  • The Saratoga Falcon · Mar 2026 On SED Talks, the teen career-speaker event I organized for the Saratoga Library Teen Advisory Board. Read ↗
  • PR Newswire · Jun 2026 National announcement of the Mott Million Dollar Challenge semifinalists, including ThinkLinks. Read ↗

07:05 · What’s next

Expanding to more schools and as many students as possible