Who we are
Kilwar is a US-based software company building infrastructure that lets businesses run AI agents safely: every action an agent takes is authorized, verified, and auditable. The product is live with early customers.
We are an AI-first company, operationally and not as a slogan. Every engineer works with Claude Code (Anthropic's CLI agent) as a standard part of the workflow. Specs, tickets, and repo conventions are written to be executed by both humans and AI agents. If this is already how you work, you will feel at home on day one. If you have never worked this way, this is probably not the right role.
The role
You will build the backend of our platform: TypeScript/Node services and SDK-based integrations that let AI agents act on real-world tools (messaging platforms, filesystems, databases) with cryptographic proof of every action. The work is Postgres, API and trust-boundary design, token verification and signing (applied cryptography, not blockchain), and the occasional product surface around it.
You will own features end to end: take a well-written brief, drive it with an AI agent doing the mechanical work, make sound design decisions inside the spec, ship with real tests, and report status honestly with evidence. "The agent says it works" is never the finish line here; you prove it works.
You should have
- 3 to 6 years of professional TypeScript/Node experience
- Deep understanding of the fundamentals: how authentication, encryption, databases, and networks actually work underneath the frameworks. You can reason about a system you have never seen
- Real, current experience developing with an AI coding agent (Claude Code, Cursor agent mode, Codex CLI, or similar). Be ready to walk us through your actual workflow on a recent task
- Postgres in production (schema design, migrations, query discipline)
- A testing habit and zero tolerance for silently swallowed errors
- Clear written English; we are an async, writing-first team
- Remote-work maturity: self-directed days with a 2 to 3 hour overlap window with the US team (late evening IST)
Nice to have
- Security or authorization experience (OAuth, tokens, RBAC, signing)
- Distributed-systems patterns: idempotency, outbox, retries done right
- Enough React/TypeScript to ship a dashboard page when the work calls for it
- Interest in Rust (our CLI tooling is Rust; a growth lane, not a requirement)
- You have written agent-facing docs (CLAUDE.md, cursor rules, agent playbooks) for a real codebase