Build the tests regulated AI has to pass.
Small team, hard problems: distributed runners, judge calibration, clinical and financial rubrics. Chennai hybrid or remote anywhere in India.
- Open roles
- 6
- Remote, India
- 2
- Chennai, hybrid
- 4
- Founded
- 2026
Open roles.
Every role lists its team, location and what you would own in the first six months.
- 1Senior Eval Infrastructure EngineerOwn the sandboxed runners, sharding and caching that keep the fast-path PR gate near 74 seconds while full suites fan out to thousands of cases.Eval InfrastructureChennai (hybrid)Eval InfrastructureChennai (hybrid)
- 2ML Research Engineer, Judge CalibrationMake LLM judges agree with physicians, attorneys and analysts, measure how much they agree, and publish the method so customers can check it.ResearchChennai (hybrid)ResearchChennai (hybrid)
- 3Clinical Evaluation LeadA physician who designs clinical golden datasets, runs annotation panels and decides what a safe agent answer looks like in prior authorization, triage and documentation.ClinicalChennai (hybrid)ClinicalChennai (hybrid)
- 4Regulatory Scientist, Financial ServicesTurn ECOA and Regulation B, FINRA Rule 2111, SR 26-2 and the EU AI Act into concrete test suites, assertions and audit exports for banking and wealth agents.Regulatory ScienceRemote, IndiaRegulatory ScienceRemote, India
- 5Developer Advocate, Agent FrameworksShow agent engineers how to put LangGraph, LlamaIndex, CrewAI and OpenAI Agents SDK agents under trajectory tests, with code that runs the first time.Developer RelationsRemote, IndiaDeveloper RelationsRemote, India
- 6Enterprise Solutions ArchitectLead technical evaluations and VPC or air-gapped deployments for banks, insurers and health systems, from first suite to signed-off production gate.SolutionsChennai (hybrid)SolutionsChennai (hybrid)
Nothing that fits? Write to careers@gonaturelle.si with something you built. We read those too.
Four weeks, five steps, no puzzles.
We test the way we want you to test agents: on realistic work, with clear criteria, and with feedback whether or not we make an offer.
- 01Week 1
Application review
The hiring manager reads it, not a keyword filter. Reply within 7 working days.
- 02Week 1 to 2
Intro call, 30 minutes
What you have built, what you want next, and the role in plain terms, including pay range.
- 03Week 2
Work sample
A paid, scoped exercise from our real backlog: a flaky shard, a judge that disagrees with physicians, a requirement to map. Four hours, your schedule.
- 04Week 3
Team loop
Walk us through your sample, then meet two future teammates and one person from another function.
- 05Week 3 to 4
Founder conversation and offer
Shiv Shankar or Rahul, then a written offer with salary, ESOPs and start date within 3 working days.
What you get.
- Salary plus ESOPsPay benchmarked to Chennai and Bengaluru product companies, with employee stock options for every full-time hire.
- Health cover for your familyGroup medical insurance of INR 10 lakh covering you, your spouse or partner, children and parents.
- Your setup, your choiceA MacBook Pro or Linux workstation and a one-time INR 50,000 home office allowance for remote and hybrid staff.
- Learning budgetINR 1 lakh a year for courses, certifications and conferences, plus paid time to publish research.
- Parental leave26 weeks for birthing parents and 12 weeks for other parents, including adoptive parents.
- Time off that gets used24 days of paid leave plus public holidays, and a company-wide week off at year end.
- Hybrid that worksChennai teams are in the Mannady office Tuesday to Thursday. Remote teammates visit once a quarter, travel paid.
Who you would work with.
Curious what the work looks like?
Run the same sandbox our customers use: a clinical suite, a trajectory gate and a calibrated judge, in your browser.