Case Study: Exotel achieves emotion detection from call audio with HackerEarth's crowdsourced ML hackathon

Crowdsourcing solution for a real-world business problem

Exotel, a cloud-based telephony platform serving over 1,300 customers and powering 3 million customer conversations daily, needed a way to extract actionable intelligence from voice calls by detecting emotions such as happiness, sadness and anger. To solve this speech‑analysis challenge, Exotel engaged HackerEarth to crowdsource machine‑learning solutions using a focused speech‑recognition hackathon on HackerEarth’s platform.

HackerEarth ran an 18‑day machine learning hackathon, supplying training audio and running submissions at scale; the effort drew 4,548 registrations, 218 teams and produced a winning emotion‑detection model. The HackerEarth program delivered high‑quality, diverse approaches that Exotel says exceeded expectations, creating a repeatable path to flagging sentiment across millions of calls and informing plans to run the event annually.

Exotel

Siddharth Ramesh

Chief Technology Officer