Senior Software Engineer, Agent Tracing

About Sentry

Software runs the world and the pace is faster than ever. Sentry helps developers fix errors and performance issues before users notice, so teams can spend less time firefighting and more time building.

Trusted by 200,000+ organizations, Sentry is today’s application monitoring standard and our team is building its AI-native future.

About the role

Agents are becoming the default way developers ship LLM-powered software, and they are miserable to debug. A single run fans out across model calls, tool calls, retries, handoffs between agents, and multi-turn conversations that only make sense as a whole. When one goes wrong, the developer usually has a log line and a shrug.

Sentry's Agent Tracing turns that into something you can actually inspect: agent runs, tool calls, model interactions, token usage and cost, conversations you can replay message by message.

As a Senior Software Engineer on the Telemetry Experience team, you'll own that product surface - Python/Django and Postgres on the backend, React and TypeScript on the frontend, sitting on top of Sentry's trace storage. You'll work with the SDK engineers who instrument the agent frameworks upstream and own everything customers touch downstream: the dashboards, the conversation views, the APIs, and the workflows that turn a confusing agent run into an obvious fix.

In this role you will

Own projects end to end. Take a problem from a rough customer complaint through scoping, design, implementation across the stack, rollout, and measurement. You'll write the design doc, cut the work into shippable pieces, and stay with it until the numbers move

Build the views developers debug agents in. Multi-agent workflows, conversation replay, cost and token breakdowns, the moment where a run failed and why. Making a deeply nested, wildly variable trace legible is a genuinely hard interface problem

Make the data underneath fast and queryable. Filtering and sorting millions of conversations, getting cost calculation right, and designing APIs that both humans and other people's agents can use well

Detect the failures worth flagging. We're expanding from showing developers what happened to telling them what went wrong - common agent failure classes surfaced as issues, not as something you have to go looking for

Take the friction out of getting started. Setup should be trivial on any stack, whether a developer wires it up by hand or asks their coding agent to do it

Collaborate daily with product managers, designers, SDK engineers, and other product engineers. You'll bring the technical perspective that makes a design better, explain constraints in terms of what they cost the user, and sweat the details other people skip: empty states, error messages, loading behavior, the edge case that only bites 2% of accounts

You’ll love this job if you

You'd rather own an outcome than a ticket. You take ambiguous problems, decide what to build, drive it across team boundaries, and follow through after launch. Nobody has to hand you a spec

You're obsessed with making an experience smooth. Friction offends you. You notice the extra click, the confusing label, the query that takes 400ms too long - and you fix them instead of filing them

You solve problems for customers, not for the codebase. Sentry's users are developers, so you can read their bug reports, sit in their tooling, and tell the difference between what they asked for and what they need

You communicate well in writing. Working across SDK teams, product, and design in a distributed company means your docs, PR descriptions, and updates do a lot of the work

You want to build in a space that's still being figured out. Nobody has settled what agent observability should look like. You'd rather help decide that than implement someone else's finished answer.

Qualifications

5+ years of professional experience building and shipping full-stack web applications

Strong experience with React and TypeScript, and with Python (Django or similar) backed by a relational database such as PostgreSQL - depth in one side and real competence in the other is fine

You've built something real with LLMs or agent frameworks and know their failure modes first-hand