- Company
- Foreverlearning
- Role
- Founding Product Designer
- Team
- Four of us, end to end
Results
- Of us built Intervoo end to end
- 4Of us built Intervoo end to end
- Pilots: the Freshworks Academy, then a week with a whole cohort at DET Hubli
- 2Pilots: the Freshworks Academy, then a week with a whole cohort at DET Hubli
- Of graduates in India are considered fluent communicators, the gap we set out to close
- 10%Of graduates in India are considered fluent communicators, the gap we set out to close
Overview
Intervoo was the first product I designed at Foreverlearning: a place to practice interviews and spoken English with an AI that actually listened and talked back.
This is how I ended up building it, and what we learned from putting it in front of students.
Where it started
A phone call I almost said no to
I was at LegalZoom when my old manager from Freshworks called out of nowhere, asking for a few hours of design help. The conversation ran long, and somewhere in it he asked if I could handle some risk. I told him if he believed in how I worked, I was in. That's how I ended up building voice-first AI products with him and three others, most of it on nights and weekends at first.
First product
Building Intervoo
The product was Intervoo, a place to practice interviews and spoken English with an AI that actually listened and talked back. Four of us built it end to end, for students who needed both interview prep and communication practice and could rarely afford a coach for either.
Competitors
Who else was in the room
Before going deeper, I went through the tools already helping people practice interviews and build resumes: Final Round AI, Huru, Interviewsby.ai, Himalayas, Apna and HackerRank. I walked through each product screen by screen, from signup to feedback, to see where they overlapped and where a student was left on their own.
How we worked
Rapid prototypes, many perspectives
We built in many rounds of rapid prototypes. Sketch it, make a clickable version, put it in front of someone, then throw away what didn't work and go again. Most screens below went through several versions before one stayed.
We were a small team and everyone was close to everyone, so every idea came with four perspectives attached. That made the discussions better, but it also meant plenty of good ideas competing for the same screen. The ones that won weren't the cleverest. They were the points that actually met the user's problem.
That became my role in the room: user advocacy. In every debate I tried to bring it back to one question, does this help the student get better? If it didn't, it didn't ship, whoever suggested it.
On the ground in Hubli
I spent more time with students and coaches than I did in Figma. We piloted with the Freshworks Academy, then a week at DET Hubli watching a whole cohort use the product: which questions landed, which felt like talking to a wall, and where a coach would have caught something the AI didn't. I also spent time comparing notes with design and product people across the ed-tech industry, including PMs from a couple of my old companies.
What the research actually told us
The numbers were blunt: only 43% of graduates in India find a job after college, and barely 10% are considered fluent communicators. That wasn't for lack of people trying to fix it: Apna, Sparrow, Ace Fluency, SpeakX and Duolingo were all circling some version of the same problem.
What none of them solved for is the part where India still leans heavily on human coaches, and for good reason. We didn't think there was a real substitute for good coaching either. The gap wasn't the coaching itself. It was that good coaches don't scale past the people already in the room with them.
What came next
What Intervoo taught us
Watching students use Intervoo settled one thing for us. An AI can listen, talk back and score an answer, but it didn't replace a good coach, and students knew it. What they were missing was access to coaches, not another app.
That's what led to TrainerTwin: a working copy of a trainer, so one expert's time can reach far more learners than their calendar allows.
