Senior Software Engineer

About Runable

Runable is one agent that runs small businesses end to end. It answers on Slack, iMessage, Telegram, Discord, Teams, web, mobile, and desktop — and it does real work: builds and deploys websites and apps, writes documents and slides, browses the internet, and executes long-running jobs in its own sandboxes.

Everyone at Runable wants to be the best in the world at what they do. That’s the bar, and it’s the real filter — not pedigree, not years, but whether being great at this craft matters more to you than being comfortable.

What you’ll build

The agent is only as good as what it can actually do, so the engineering runs deep:

• The agent itself. Orchestration across frontier models, tool use, long-running and resumable work, and the prompt and evaluation systems that decide whether the agent is trustworthy enough to run someone’s business.
• Execution and hosting infrastructure. Sandboxes where the agent builds, browses, and deploys — and the hosting behind what it creates: websites, apps, databases. Infrastructure at scale that has to be genuinely cheap per customer. Making a unit of agent work cost cents instead of dollars is a core engineering problem here, not a finance afterthought.
• The product surface. Web, mobile, desktop, and every messaging channel a business owner already lives in — one TypeScript monorepo on Postgres, Redis, and DBOS for durable execution, with Cloudflare and GCP underneath.

The stack is deliberately the least interesting part of that list. Tools change; the problems don’t: durable execution, cost at scale, an agent you can trust with a business.

What’s different about working here

• Product depth is an engineering value. Engineers here understand the business the agent is running, not just the service they own. The best technical decisions come from knowing why a store owner churns, what a failed deploy costs them, and what “done” means to someone who isn’t technical.
• You own outcomes, not components. Not “the queue service” but “the agent can take a business from zero to live”: design, infrastructure, cost, and the proof that it works.
• Builders all the way up. A company of individual contributors. Nobody manages as their job; the people setting direction are also shipping. If you want impact measured by what you build rather than the size of your team, you’ll fit right in.
• Agents write a lot of our code. The instructions, skills, and runbooks that make them good live in the repo and are maintained like code. You’ll get very good at directing AI — and at knowing when not to.
• Small diffs win. The change that deletes or reuses code beats the one that adds it. Speculative abstraction gets rejected in review. The repo is the source of truth, not tribal knowledge.

What we care about most: focus

The quality we weight above everything else in this role is focus. Engineering work here fails when it becomes a tour of shiny systems, half-finished rewrites, or model demos that never become dependable product. It compounds when one person takes a hard problem, owns it completely, and keeps going through the boring middle until the agent is meaningfully more reliable, cheaper, faster, or more useful for the business owner using it.

If your instinct when a system is flaky, expensive, or underspecified is to keep digging until it is actually solved, this is your role. We judge you on the depth and durability of what you build, not the number of surfaces you touched.

Who we’re looking for

• 2+ years building production systems — but the real question is whether you want to be the best in the world at this. If that reads as marketing copy, we’re probably not your place. If it reads as a relief, keep going.
• You’ve run something real in production and cared what it cost. You’ve made the graph go down, not just up.
• Product sense with depth: you ask what the user was trying to do before you ask what the stack trace says.
• High agency. You notice what’s broken, fix it, and tell people — without being asked.
• You want to work with AI agents at the frontier: as the product, as the user, and as a collaborator in your editor.

Hiring process

Two steps, no gauntlet:

1. A 15-minute introduction call with Saksham, co-founder.
2. A work-day trial in our office — you work on a real problem with the team, and both sides find out what it’s actually like to work together.

Location: In person, Bangalore.