About KIM

KIM is an AI-native services company building AI-powered customer support for D2C e-commerce brands — work done by AI, vetted by humans. Our core AI system handles real customer conversations for real brands, every day. It's not a demo. When it gets something wrong, a customer notices.

The Role

You'll work directly alongside our senior AI engineer on our hardest AI problems. This role leans more toward applied research than conventional backend engineering — you'll spend your time on questions like:

- How do we know a system change actually made responses better? (Eval design, harness construction, failure taxonomies)
- Where should reasoning be deterministic vs. LLM-driven?
- Why did retrieval miss, and how do we fix the class of failure rather than the instance?
- How should an AI system remember things over long-running relationships?

You'll design experiments, build eval datasets that reflect real-world distributions (not cherry-picked queries), diagnose regressions, and argue for architectural decisions with evidence.

Who We're Looking For

- AI-forward. You follow the field closely because you want to, not because you have to. You've formed opinions about what works and what's hype.
- Strong intuition for modern AI and backend systems. You understand LLM pipelines, retrieval systems, and agent architectures well enough to reason about their failure modes — and enough backend fundamentals to ship what you design.
- Comfortable in ambiguity. Problems here rarely arrive well-specified.



You can take a vague concern ("the recent version feels worse") and turn it into a measurable question.
- Thinks in abstractions. You naturally lift a problem to the right level — you see the class of failure, not just the instance — and you have taste and conviction about how things should be built.
- Pushes back. You disagree openly, argue from first principles, and change your mind when the evidence says so. We want proper discussions, not silent agreement.
- Builds for fun. You have personal projects — things you made because you were curious. We'd genuinely like to see them.

Great to Have
- Hands-on experience with evals — building harnesses, designing datasets, measuring LLM output quality, catching eval inflation and dataset contamination.
- Experience with RAG systems in production: retrieval debugging, reranking, chunking strategy.
- Familiarity with agentic patterns (ReAct-style loops, tool use, orchestration) and where they break.
- Prior work at an AI-first startup, research lab, or serious open-source AI contributions.

What You'll Get
- A seat next to the people making the core architectural decisions — small team, no layers, direct access to founders.
- Real production feedback loops: your experiments ship to live customer conversations, and you see the results in days, not quarters.
- A research-flavored role with the accountability of a product company — the best of both.

Drop your CV at [email protected]

📌 Artificial Intelligence Engineer (Bengaluru)
🏢 kim.cc
📍 Bengaluru