Staff Data Scientist

Netradyne builds camera based fleet safety systems, with on device computer vision that scores driver behaviour and more than 25 billion miles of video analysed. This role is not core model development, and the posting says so directly. It is the evaluation and analytics seat: owning the frameworks, metrics and audit methodologies that decide whether the AI features actually work in real world deployments. You would design offline and online evaluation frameworks, define and evolve the KPIs used across teams, run deep error and bias analysis to find failure modes, and own golden datasets and benchmarking standards. There is real statistical rigour expected, including designing and reviewing A/B tests and controlled rollouts and getting confidence intervals and significance right. The posting also expects you to challenge existing metrics when they fail to capture ground reality, which is the interesting part of the job.

What the posting requires as mandatory:
- A B.Tech, M.Tech or PhD in Computer Science, Electrical Engineering, Statistics, Mathematics or a related field.
- 5+ years of experience in data science, analytics or a closely related domain.
- Strong foundation in probability, statistics and estimation theory.
- Strong programming in Python, with solid fundamentals in object oriented programming, algorithms and data structures.
- Deep familiarity with SQL, complex query writing, indexing and database internals, plus working knowledge of at least one NoSQL store.
- Experience with data visualisation and analytical storytelling.
- Excellent written and verbal communication.
- Familiarity with AI powered tools for analytics and software development, including using them for exploratory data analysis, experiment analysis and documentation, and critically evaluating AI generated output for correctness, statistical validity and reproducibility.

Preferred skills:
- Exposure to cloud platforms and services such as AWS Kinesis, EKS and autoscaling systems.
- Experience building lightweight web or service components with Flask or Django.

What the work actually looks like:
- Design, implement and maintain offline and online evaluation frameworks for AI and ML features.
- Define, validate and evolve the KPIs, success metrics and audit methodologies used across teams.
- Perform deep error analysis, bias analysis and segmentation to identify failure modes.
- Own golden datasets, validation protocols and benchmarking standards.
- Conduct large scale analytical studies on feature performance, data quality and system behaviour.
- Design and review experiments including offline evaluations, controlled rollouts and A/B tests, ensuring statistical correctness.
- Perform post deployment monitoring and regression detection.
- Build tools, dashboards and automation to scale audits and reporting, and enable self serve analytics.
- Mentor junior data scientists on statistical rigour, experiment design and analytical storytelling.

Location and working pattern: Bangalore. This posting does not state the number of office days, though some other Netradyne listings specify a five day on site model, so ask.

A note on the employer: Netradyne postings carry a recruitment fraud warning. Official communication comes only from netradyne.com or us-greenhouse-mail.io addresses, and the company never charges candidates a fee.

Honest fit guidance: if you want to build models, this is explicitly not that role. It suits someone who is happier proving whether a system works than building it, and who can tell a product team its metric is wrong.