About the role
The Agents team runs the agents that do the coordination inside Clexerl Go. An agent reads a person's calendar, drafts a follow-up to a client, moves a task and asks for approval, and the runtime has to do this reliably, within a token and latency budget, and with a clear record of every step. The problem is less the model call than everything around it: state, tools, retries and cost.
The runtime is the layer where Frontier AI's models meet Clexerl's products. Improvements to prompt caching, speculative decoding on the serving path or long-term memory retrieval show up here as lower cost per task. Failures found by the safety evaluations come back as runtime guardrails, so a bad tool call is caught before it touches a person's data.
What you’ll do
- Build the agent executor that runs plans as durable steps, with checkpoints, timeouts and resumable state.
- Implement tool-calling adapters for calendars, tasks and CRM actions, each with schema validation and permission checks.
- Design per-run budgets for tokens, tool calls and wall-clock time, and enforce them in the scheduler.
- Add tracing that links each model call, tool invocation and human approval into one inspectable run record.
- Tune the serving path with KV-cache reuse, batching and speculative decoding, and measure each change against a fixed eval set.
- Write regression suites from real failed runs, and block releases that move the eval scores the wrong way.
What you’ll bring
- Experience building backend services in Python, Go or TypeScript, with queues, retries and idempotent writes.
- Practical understanding of large language model APIs, tool use and the failure modes of structured output.
- Comfort with distributed tracing, cost metrics and profiling latency across model and non-model components.
- Experience operating services on Kubernetes or a similar scheduler, with on-call and incident follow-through.
- A habit of measuring before optimising, and of writing down what an experiment proved and did not prove.
Nice to have
- Experience with GPU inference servers, quantisation or batching strategies for transformer serving.
- Background in workflow engines, durable execution or event-sourced state machines.
- Familiarity with prompt injection, sandboxing of tool execution or other agent security patterns.
- Work on evaluation harnesses, offline replay or statistical testing of non-deterministic systems.
Work you might take on
- 01
Build a durable execution layer that resumes an agent plan after a worker crash without repeating completed side effects.
- 02
Add a per-tenant token budget with soft warnings and hard stops, surfaced in the run trace.
- 03
Prototype speculative decoding for the planner model and report the tokens-per-second gain against quality.
- 04
Create a replay tool that reruns last week's failed agent runs against a candidate model and summarises the differences.
Compensation
$230K – $360K
The US base salary range for this role. Every offer also includes equity in Clexerl and, when it spins out, in the venture you help build. We tell you the reasoning behind your number.
Benefits
Health, fully covered
Medical, dental and vision for you and your dependants, premiums paid in full.
Two kinds of equity
Equity in Clexerl, plus a stake in the venture you help build when it spins out.
22 weeks parental leave
Fully paid, for every parent, however your family arrives.
Rest that sticks
Flexible time off with a four-week minimum we actually check on, plus a company-wide week in December.
Learning and compute
A yearly budget for courses, conferences and books, and compute credits for side projects.
401(k) with a 4% match
Matched from your first paycheque and vested immediately.
How we hire
Apply 5 minutes
Send links to what you have built and a few lines on why this role. No cover letter needed.
Intro call 30 minutes
A conversation with the hiring manager about the work, the team and what you want next.
Work sample 2–4 hours
A paid take-home or a pairing session on a real-shaped problem. You pick the format.
Studio day One day
Meet the team, go deep on your work and ours. Remote or in person.
Offer Within 5 days
A written offer with the full range, the equity and the reasoning behind both.
Opens an email with a short template. Links beat résumés.