Twitxy Saved

bolna

Verified account This account is verified. Learn more

@bolna_dev

Powering Voice AI for India (@ycombinator f25)
🌍 bolna.ai Joined January 2024
30 Following 1.4K Followers 410 Posts

Latest posts 5

Pinned post

bolna Verified account @bolna_dev
Voice, innovation, and ideas are about to align in perfect harmony under one roof!

Take your seats for the biggest orchestration ever... this November, in Bengaluru.

Symphony is coming alive! Stay tuned!
0:08
1 replies 6 reposts 20 likes 2.9K views View on X
bolna Verified account @bolna_dev
Happy Diwali from Team @bolna_dev! 🪔

Last Diwali, five of us celebrated around a single table. This year, we needed a considerably larger room for a full house (the group photo now needs a wide-angle lens).

Slack went quiet for an evening, and nobody missed it.

Wishing your Diwali is as bright and happy as ours was! ✨
0 replies 3 reposts 17 likes 691 views View on X
bolna Verified account @bolna_dev
Voice Agents do more than understand what you say. They decide what to do next.

Introducing DECIDE, a benchmark for evaluating those decisions on real production calls.

benchmark.tazolabs.com
benchmark.tazolabs.com DECIDE · A benchmark for voice-agent decisions 4,867 decision questions. 1,081 production-call clips. 32 tasks. A benchmark for deciding what happens next in a voice conversation.
Naman Jain Verified account @namanjn1998 2d
Introducing DECIDE: a benchmark for audio decisions in voice agents, in association with @bolna_dev

A voice agent doesn’t just transcribe. It decides: speak or wait, route the call, pick a tool, switch language. Text can’t tell you if the caller sounded annoyed, interrupted, or talked over the agent.

DECIDE tests exactly that, on real production calls.

→ 4,867 decision questions from 1,081 real agent-caller clips (3 to 25 seconds)
→ 32 tasks across 6 families: turn-taking, graph routing, tool decisions, language switching, caller emotion, business intent

Direct-audio models and STT + text models (20 in total) are scored on the same questions, so you can see what audio actually adds, and what it costs in dollars and milliseconds.

Submit your model endpoint and get scored; no invite needed.

Website: benchmark.tazolabs.com
Samples: huggingface.co/datasets/tazol…
View quoted post on X
0 replies 6 reposts 20 likes 503 views View on X
bolna Verified account @bolna_dev
Bengaluru, it’s time to plan your day! 🚀

The agenda for the big day, Symphony: India’s flagship Voice AI Summit, is finally out!

Across Sur and Abhyas, the day runs on two stages at once. Sur, our main stage, brings together the people shaping voice AI in India: global AI labs, founders building speech models for India’s languages, and the investors betting on what comes next. Abhyas, the build stage, is for those who want to see how it all comes together, with product showcases, sessions and the Buildathon finals.

Team @bolna_dev will also be unveiling what our team has been quietly building for months, so do stay for that.

Grab a coffee, pick your sessions, and maybe pack your comfy shoes for running between stages. Link to the full agenda: bolna.ai/symphony/agenda

P.S. You’ll spot a special performance on the agenda, and it’s the part Team Bolna is most excited about. Keeping it under wraps has been killing us, so stay tuned.
bolna Verified account @bolna_dev
Enterprises that answer calls across Ladakh to Kanyakumari or Gujarat to Manipur shouldn't be limited to agents who only know how to answer in one language. Whether it's a bank, a healthcare facility or a national university, the caller on the line deserves an agent that meets them where they are.
The solve? A dedicated prompt for every language.

Except that always becomes too tedious to build with each add-on.
Adapting to voice agents was supposed to make this easier. Instead, adding new languages only meant adding another version of the agent with an additional prompt to write and test, with each linguistic version drifting apart and getting lost in translation.

@bolna_dev's latest update on Canvas takes away that friction for enterprises with just one edit.
Draft your agent’s logic once, and it holds across every language the agent is supposed to speak. Only the lines the agent says out loud change from one language to the next. Bolna translates them automatically, and teams can still fine-tune the wording so it sounds natural to local callers.

Adding a new language takes just a click, and not a rewrite. What you end up with is multiple agents fluent in the language your customers speak without an added effort of a dedicated prompt for each one of them.

Go try your new Canvas on Bolna now!
1:14