Software Engineer, Agents

In 2026, we're on the verge of an "infinite software crisis." How will we support, maintain, and operate the explosion in application development?

Sazabi gives teams a single place to ask questions about their production systems in plain language, automatically visualize what's happening, and get to the root cause 10x faster. No tedious instrumentation. No dashboard configuration. No alert tuning. Just answers.

We're backed by tastemakers from the world's top AI companies: Vercel, Graphite, Daytona, Browserbase, LangChain, Mastra, Replit, and more.

Build and iterate on the core AI agents that power Sazabi

Design systems for anomaly detection, root cause analysis, and automated debugging

Work on prompt engineering, tool use, and agent orchestration

Improve reliability, latency, and correctness of AI-driven workflows

Experiment rapidly with new models, frameworks, and techniques

Translate messy real-world production issues into structured AI workflows

Strong experience working with LLMs in production (agents, RAG, tool use, etc.)

Deep curiosity about how and why AI systems fail—and how to fix them

Ability to prototype quickly and iterate based on real-world feedback

Strong engineering fundamentals (this is not just prompt hacking)

Comfort operating in a fast-moving, experimental environment

Bonus: experience with observability, debugging systems, or developer tools

Our stack is TypeScript end-to-end. We use PostgreSQL for relational data, Temporal for durable execution, and the Vercel AI SDK to orchestrate our AI agents against the latest models. At the infrastructure layer, we use Terraform, Kubernetes, and AWS.

Some of the most interesting engineering problems we grapple with are:

AI agents at production scale

Log processing at scale

Multi-region data architecture

Real-time streaming infrastructure

Evals and reinforcement learning

Zero-downtime deployments and fast rollbacks

Our team comes from Brex, Y Combinator, Rootly, Google, Doppel, and other top tech companies. We move fast and hold a high bar for both the product and each other.

We operate according to six values:

Integrity. We say what we mean, do what we say, and own our mistakes.

Urgency. We treat every week like it matters, because it does.

Craftsmanship. We care deeply about the quality of our work.

Service. We're here to make developers’ and our teammates’ lives easier.

Kindness. We're direct, and we move fast, but we treat each other well.

Fun. We genuinely enjoy working together. If you don't love what you're building and who you’re building it with, what's the point?

You can find out more about us here:

Watch our launch video

Read the manifesto

Follow us on X and LinkedIn

Sazabi is more than just a tool. It’s also a philosophy. We're taking a radically different approach to observability, centered on three big ideas:

Less is more. Most observability platforms drown you in dashboards and modules that you don’t need. We believe observability needs less UI, not more. Sazabi surfaces exactly the information you need in a simple, beautiful chat interface.

Logs are all you need. The “three pillars of observability” idea is outdated. Sazabi accepts only one kind of telemetry: logs. This simplifies the instrumentation and product experience dramatically.

Monitoring is dead. Creating and maintaining static alerts for a fast-evolving system is a fool’s errand. The future is agentic anomaly detection: AI agents that automatically check your app for issues 24-7.

We came to these beliefs the hard way — after years of building infrastructure and responding to incidents with tools that offered plenty of bells and whistles, but no clear answers when it counted most.