Software Engineer — AI Infra
About the role
Tasklet is building an AI platform for knowledge work. The role focuses on the core systems that power Tasklet’s agents, including agent harness, context management, filesystem, model serving, databases, Kubernetes, networking, and cloud infrastructure. It is a hands-on IC role requiring deep CS and systems fundamentals, with flexibility to work across runtimes, storage, distributed systems, and production infrastructure.
Responsibilities
- Build the agent harness, including orchestration, tool execution, state, evaluation, and failure recovery.
- Develop context-management, filesystem, memory, storage, and sandboxed-execution systems.
- Work on model routing and serving, databases, Kubernetes, networking, queues, and cloud infrastructure.
- Make long-running agent workloads faster, safer, more reliable, and more cost-efficient.
- Turn fast-moving advances in models and agent systems into robust production capabilities.
- Work across the stack and across role boundaries whenever that is the best way to solve the problem.
Qualifications
- You’re an exceptional problem solver with deep computer-science and systems fundamentals.
- You reason from first principles, form precise mental models, and enjoy understanding what is happening all the way down the stack.
- You’ve built and operated complex production systems and can debug across application code, runtimes, storage, networking, and infrastructure.
- You care about elegant architecture and real operational outcomes: performance, reliability, safety, and cost.
- You learn quickly, operate well with ambiguity, and prefer responsibility over process.
- Prior professional AI experience is not required, but genuine depth of engagement is expected. You already use current models seriously, probe their limits, follow the field closely, and have your own ideas about how AI systems should work.
Skills
Benefits
- medical, dental, vision
- 401(k)
- four weeks of PTO
- free lunch
- and more
About Tasklet
The cloud agent operating system for knowledge work
AI chatbots can answer questions, but they can't do your work — they aren't connected to your tools, can't take actions, and stop running when you close the tab. Meanwhile, knowledge workers are still spending most of their day switching between dozens of SaaS apps, manually stitching together workflows that should be automated. Tasklet is the AI agent that actually does the work. It connects to every tool you use — via thousands of built-in integrations, any HTTP API, or MCP servers — and takes real actions on your behalf: pulling reports, updating CRMs, triaging emails, processing data, and more. When it needs a UI, it generates one on the fly. When it needs to crunch numbers or do bulk operations, it writes and runs code in a cloud sandbox. When it needs to interact with a website that doesn't have an API, it uses a real browser. And because it runs in the cloud 24/7, you can set up triggers to have it take ongoing responsibility for workflows — not just do something once, but handle it going forward. The product is deliberately simple: a chat interface where you talk to an extremely capable agent. But underneath that simplicity is a cloud-native agent runtime, best-in-class context engineering, a universal connections layer, and multi-model support that keeps us on the frontier as AI improves. We think legacy SaaS is being unbundled by agents, and Tasklet is the operating system that ties it all together.
More open roles at Tasklet
- Head of Design Engineering2mo agoHybrid (San Francisco, CA)
- Software Engineer — Platform2mo agoHybrid (San Francisco, CA) · Engineering
- Software Engineer — Product2mo agoHybrid (San Francisco, CA) · Engineering
- Customer Success Engineer3mo agoSan Francisco