Build in public

Our company diary.
Open to everyone.

What we are working on, what we shipped, where we got it wrong, and what we learned. Posted here every week, not just on social media. You can come back to any month and see exactly what was going on at Kuvaka.

Tell us what to build next

Kuvaka at the Madhya Pradesh Tech Growth Conclave 3.0

Kuvaka started in Bhopal. This week, we're back in Madhya Pradesh — now operating out of Indore — with a booth at the Madhya Pradesh Tech Growth Conclave 3.0, showing two products we've spent the last several months building in the open.

We had to make a call on what to bring, and we didn't bring everything. Decrackle is on full display — it's the most thoroughly benchmarked thing we've built, and a state-level tech pavilion is exactly the audience that cares about hitting a 2.04% word error rate, +10.6dB SI-SDR gain, and 28–32x real-time speed, fully self-hosted. From the HealthTech suite, we're showing Medical History Taker and Clinic Pro.

Kuvaka's Note Taker app — built directly on Decrackle's audio AI stack — is going to record and transcribe conversations at the booth, live, in real time. Stop by and talk to Shantanu or Kartik, and you'll watch your own conversation get transcribed on screen while you're having it.

Read full post

Update28 June 2026

The new Kuvaka website is live.

We launched the first version of the new kuvaka.io today.

It is not perfect. There are things we will change, content we will add, and sections that will improve as we build in public and hear from people who use it. But it is honest, it is ours, and it reflects how we actually think about what we are building.

New positioning. Three industries. Open-source products you can use for free today. A pricing page with real numbers. And this — a place where we will keep talking, not just post announcements.

If something is unclear, missing, or wrong — tell us.

Event28 June 2026

GITEX AI Europe 2026

Messe Berlin, Germany

Kuvaka is going to GITEX AI Europe 2026 in Berlin. Come find us.

30 June – 1 July · Messe Berlin, Germany

We will be at GITEX AI Europe 2026 in Berlin this week — and we want to meet the right people while we are there.

GITEX AI Europe brings together AI pioneers, enterprise leaders, frontier startups, investors, and policymakers to drive real-world AI deployment and cross-border collaboration across Europe. For us, it is the right room at the right time.

We are an AI-native software company with roots in India and a deliberate move into Europe. We build open-source AI solutions for healthcare, manufacturing, and enterprise — and we deliver customisation through senior pods in fixed 7-day sprints. Our new positioning, our new website, and our European presence are all live as of this week.

GITEX Berlin is where we want to start those conversations in person. With European enterprises looking to digitise. With healthcare organisations evaluating AI for clinical workflows. With manufacturers who need agentic operations. With investors who understand what it means to build an open-source-first AI company with a real delivery model.

What we want to talk about at GITEX

Our HealthTech open-source shelf and what we are building next. The pod model and how 7-day sprints work in practice. European SME digitisation and what that actually looks like end to end. Any problem you think AI should already be solving.

If you are attending and you want to talk — book a slot with our team directly. We will be there both days.

We will update this post with what we learned, who we met, and what came of it once the team is back.

Book a meeting with our team at GITEX

News28 June 2026

Kuvaka is applying to the IndiaAI Startups Global Acceleration Program — 2026, Cohort 2.

A government-backed program to take Indian AI companies global. We think we belong in that room.

We have applied to the IndiaAI Startups Global Acceleration Program, Cohort 2 — run by the IndiaAI Mission under India's Ministry of Electronics and Information Technology, in collaboration with Station F, the world's largest startup campus, and HEC Paris.

The program is designed to help high-potential Indian AI startups expand into global markets — specifically Europe — through mentorship, investor access, and strategic partnerships.

Why we applied

Kuvaka is already moving into Europe. We have a registered entity in Germany, a clear focus on European healthcare and Mittelstand SMEs, and a delivery model built for the European market — transparent pricing, open-source-first, sprint-based, GDPR-aware.

But being in Europe and being established in Europe are two different things. This program offers access to 1,000+ AI experts and 500+ investors from Station F and HEC Paris — exactly the network that accelerates the second part.

We are also a healthcare AI company. The second cohort specifically spans HealthTech, among other areas — which aligns directly with what we are building. Our History Taker product, our Care Notes solution, and our open-source HealthTech shelf are all live or in active development. We have already done primary research with practising physicians. The work is real.

What we are hoping for

Honest answer: validation, network, and acceleration. We are not applying because we need the program to tell us what to build. We are applying because the right introductions, in the right European ecosystem, at this specific moment in our growth, could meaningfully compress timelines.

We will share the outcome here — selected or not.

If you are connected to the IndiaAI Mission, Station F, or HEC Paris and want to learn more about what we are building — reach out.

Talk to our team

ProductHealthTech28 June 2026

We talked to doctors before we wrote a single line of code. Here is what they told us.

How real conversations with physicians shaped History Taker — and why we think we are building this the right way.

The first version of History Taker was not designed in a product sprint. It was designed in a series of conversations with doctors — practising physicians who see patients every day, manage overwhelming documentation loads, and have spent years working around tools that were built for the system, not for them.

This is one of those conversations. We are sharing it because we believe the research behind a product matters as much as the product itself. Building in public means showing the thinking, not just the outcome.

What we went in asking

We did not go into these conversations with a solution. We went in with a problem statement — that clinical history-taking is time-consuming, inconsistent, and largely manual — and we asked doctors to tell us whether that matched their reality, and what the shape of their actual pain looked like.

What we heard

Several things came up consistently across the physicians we spoke to.

The documentation burden is happening at the wrong time. Doctors are writing notes during the consultation, or after it, when they should be focused on the patient. The cognitive split between listening and recording is real and costly — both in clinical quality and in the doctor's time and energy.

History is inconsistent across patients and clinicians. Without a structured intake process, history quality depends on the individual clinician's style, the patient's ability to recall and communicate, and the time available. The same condition in two different patients can generate completely different records, making comparison, audit, and handover harder than it needs to be.

Doctors do not want to be replaced. They want to be supported. Every physician we spoke to was clear on this. They are not looking for AI to make the clinical decision. They are looking for AI to handle the capture, the structure, and the prompting — so they can focus on the judgment that only they can exercise.

Missed information is a clinical risk, not just an admin problem. When a history is incomplete, it can lead to missed diagnoses, incorrect prescriptions, or avoidable readmissions. A tool that prompts for missing information in real time is not a nice-to-have.

What we are building, and why

Based on what we heard, History Taker works in two modes.

In the manual-assist mode, the doctor conducts the consultation in their usual way. History Taker listens, structures what it hears, and — critically — alerts the doctor in real time if something in the history appears incomplete or inconsistent with the department protocol. It does not interrupt. It surfaces a flag. The doctor decides whether to act on it.

In the AI-led mode, History Taker conducts the intake conversation with the patient before the doctor enters the room. It adapts its questioning to the department the patient has been admitted to — a general ward admission triggers a different question set than a cardiology intake — and delivers a structured, complete history summary ready for the physician to review and build on.

In both modes, all data is encrypted end-to-end and the system is built to HIPAA standards from the ground up. The record it produces is a clean, structured repository — not a transcript — that becomes the foundation for the consultation and the patient record.

After the history is taken, History Taker generates a summary with suggested diagnostic directions and, where relevant, prescription suggestions keyed to medicine availability in the facility — for the doctor to approve, modify, or overwrite. The doctor always has the final word. The AI presents options, not decisions.

Why we are sharing this now

Because this is how we think software should be built. Not from assumptions, not from competitive benchmarking, not from what a product manager thinks a doctor needs. From the people who will use it, before a single feature is scoped.

We have more of these conversations recorded. We will share them here as the product develops.

If you are a physician, a clinical informaticist, or a healthcare organisation and you want to tell us what we are getting right or wrong — please do.

Give us feedback on History Taker Explore on GitHub Book a HealthTech discovery call

This is our public diary. We post here every week — wins, failures, decisions, and work in progress. If you want to tell us what to build next or what we should improve, the feedback form is always open.

Tell us what to build

Follow updates