Applied Data Scientist

## Overview

Self-driving companies record every mile a car drives. Nobody records how it gets fixed. Repair expertise lives in technicians' heads: undocumented, unstructured, and disappearing as their shortage grows.

Bayworks captures voice and video data as technicians repair cars and verifies the work. Every repair becomes ground truth for how cars get fixed: symptom, diagnosis, fix, outcome. This compounds into a system that runs the dealership's operations and standardizes every repair, making it faster and compliant with manufacturer specifications.

## Why now

In 110 days, we've gone from 0 to 80+ dealerships covering 4,000 cars a day, including a top-50 US dealer group ($3B+ annual revenue) and have hit $XMn ARR. We know what's working.

Now we need to get to 1000 rooftops / $10M ARR. To get there we need someone to own our data: turning every repair we capture into ground truth, and building the models that run on it. The data is there; you turn it into the product.

## World-class team

We're engineers from BITS Pilani.

* **Kaustubh (Founder & CEO)** built voice AI at Sarvam (India's frontier AI lab) deployed across government and public enterprises, and before that worked in automotive retail lending with carmakers.
* **Shubham (Founder & CTO)** built the underwriting engine at Navi, a $2B fintech, powering $10M+ in loans every day, and the mobile app at Scapia, which scaled to millions of users.

## Investors

Backed by Y Combinator (F26)

## What you'll work on

* Turn raw voice, video and dealership data from every repair into structured ground truth: symptom, diagnosis, fix, outcome. This is a hands-on role; what you build ships to production
* Build the evals that tell us whether our AI got a repair right, and own the metrics we ship against
* Run experiments across speech and language models and decide what goes to production
* Build models on repair data that standardize the next repair: the likely diagnosis for a symptom, and what a fix needs to meet manufacturer specifications
* Spend time in service bays with technicians and feed back what you learn into the product

## You can be a great fit if

* You've shipped ML or data products to production, not just notebooks
* You're comfortable with messy real-world data: noisy shop-floor audio, free-text repair notes, inconsistent dealership records
* You're strong in Python and SQL, and have built evals for LLMs, speech or multimodal models
* You move fast and don't overthink
* You care whether the model works in the service bay, not just on the benchmark
* You want to be at a company that's early enough that your work actually matters

## Compensation

* **This role includes:**
* 160K - 200K base salary
* Meaningful equity
* Medical, dental, and vision insurance
* Free DoorDash vouchers for everyday meals

## About the interview

1. 30 mins with the founders
2. Assignment: a short take-home on a sample of real (anonymized) repair data
3. Offer