The posting
ABOUT BOLNA
Bolna is Voice AI infrastructure built for India, and now for the world. We help businesses deploy intelligent voice agents that can call, converse, and convert in any language, at scale. From collections to customer support to sales, our agents handle millions of conversations so humans don't have to.
We're a YC F25 company, backed by General Catalyst, with 1,050+ paying customers and growing fast. Our team of 70+ is based in Bengaluru.
THE ROLE
Bolna's voice agents handle millions of real conversations for some of India's largest enterprises. Every one of those calls has to sound natural, and when one goes wrong, we need to know why, fast.
As a Product Manager, you'll own two areas: how the call sounds and feels to the person on the other end, and how we, our FDEs and our customers know whether an agent is working and why it isn't. You'll decide which problems matter most, define what "better" means, and ship improvements that work for every customer, not just one. This role is for someone who has done this before: you'll set the direction for both areas, not execute a roadmap someone else wrote.
You'll work with the founders and the engineering, ML and FDE teams, sometimes on your own and sometimes alongside another PM or an engineering lead. It's a hands-on role: you'll write the spec, and often the first SQL query, prompt or prototype too.
WHAT YOU'LL DO
IN-CALL EXPERIENCE
- Own the quality of the live conversation: background noise and noise cancellation, turn-taking and interruptions, latency, and language switching.
- Drive agent memory: context that carries across calls, between steps of a multi-step agent, and through agent-to-agent transfers.
- Work with ML engineers on model choices and the models we build. Define what "better" means, build the eval set, and decide when an in-house model is ready to replace a vendor.
- Turn fuzzy complaints from FDEs and customers ("the agent sounds off", "it keeps cutting people off") into metrics with clear thresholds.
OBSERVABILITY
- Make it fast to find out why a call went wrong: which turn, which component (telephony, speech recognition, LLM, voice, tools), and what to change.
- Shape the dashboards, call-level debugging views and automated evals, including LLM-as-judge, that FDEs and customers use every day.
- Define the handful of quality and reliability metrics the company watches, and own their accuracy.
- Spot recurring failure patterns across thousands of calls before customers report them.
WHAT WE'RE LOOKING FOR
MUST-HAVE
- 4+ years in product management, or in an engineering, consulting, analytics or founding role where you owned what got built and why. You've shipped products that moved a metric you can name.
- You communicate clearly in writing and in person. A one-page problem statement from you gets engineers, FDEs and customers on the same page.
- You're a driver with high ownership. You take responsibility for outcomes, not tickets, and follow up until the fix works in production.
- You're comfortable handling many small problems in a week and one hard problem over a quarter.
- You can work independently, and you also work well alongside other PMs and engineering leads.
PREFERRED
- Problem solving. You break messy, ambiguous problems into clear parts, find the root cause, and pick the simplest fix that works.
- Data sense. You go to the data before forming a view. You write your own SQL, can tell a real pattern from noise, and notice when a number looks wrong and work out why.
- Technical understanding. You know how systems fit together and can reason about LLMs, APIs, latency and trade-offs. You read logs and API docs and hold your own in design discussions with engineers. You don't need to write production code.
GOOD-TO-HAVE
- Experience with voice AI, conversational AI, LLM products, or developer and infrastructure platforms.
- Hands-on with LLMs: prompting, evals, LLM-as-judge, or prototypes you vibe-coded yourself.
- Exposure to speech technology: speech recognition, text-to-speech, noise suppression, or telephony.
- You've worked closely with enterprise customers or with forward-deployed or solutions teams.
WHAT SUCCESS LOOKS LIKE
In your first year, success means:
- Callers on our largest deployments have measurably better conversations, tracked on metrics you defined.
- FDEs and customers can work out why a call went wrong in minutes, without pulling in engineering.
- Recurring call failures are caught before customers report them.
- Every in-call change ships with a metric, so we know whether it worked in production.
- You own the roadmap for both areas, so engineering time goes to the problems that matter most to customers, and everyone can see why.
WHY BOLNA
- One of the fastest-moving teams in Voice AI in India
- Your problems come straight from live deployments at some of India's largest enterprises, and your fixes reach millions of calls
- Direct access to the founders, and daily work with engineering, ML and FDEs
- Build a rare skill set at the intersection of AI, speech, product and enterprise
- Competitive compensation + meaningful ESOPs
- Bengaluru office, in-person team collaboration
