Time since we first saw it
Day 3
Similar jobs usually come down between day 1 and day 15.
From 1,371 similar postings that came down. Coming down is not always a hire.
United States · Latin America · Canada · Europe · California, United States
Actively hiring, first seen by us 3 days agoLocation type: RemoteFrom the employer’s posting
Employment type: Full-timeFrom the employer’s posting
Department: EngineeringFrom the employer’s posting
Own the AI automation project: architecture, reliability, observability, cost and roadmap
Build eval and benchmarking setup covering real customer workflows, with regression checks
Push run success rates up and add recovery that fixes runs instead of failing them
Bring cost per run down through model routing, caching and provider choice
Build end-to-end tracing so failed runs can be replayed and diagnosed quickly
You own the AI automation project end to end: architecture, reliability, observability, cost, and the roadmap of what it can do next.
Build an eval and benchmarking setup that covers our real customer workflows, with regression checks that run on every change.
Push run success rates up across changing UIs, sessions and edge cases, with recovery that fixes the run instead of failing it.
Bring cost per run down with model routing, caching and provider choice, without losing accuracy.
Build end-to-end tracing and observability, so any failed run can be replayed and diagnosed in minutes.
Actively hiringFirst seen by us 3 days ago, still listed, and not a repost of an earlier posting.
These dates show when we found the posting. Check the employer’s website to confirm it is still accepting applications.
Day 3
Similar jobs usually come down between day 1 and day 15.
From 1,371 similar postings that came down. Coming down is not always a hire.
2 open jobs
Open jobs by the week we first saw them, last 12 weeks.
Hiring recordAdd your resume and we’ll mark which of these tools you already use, and point out anything in the posting that could rule you out.
- Experience managing LLM cost and latency in production: model routing, caching, provider choice. - Strong in TypeScript and Python. You'll work in both. - Architecture judgment: you can take a system that was built fast and set the direction for its next stage.
- Experience with model gateways and providers such as OpenRouter or Bedrock. - Familiarity with Django and ClickHouse, or similar stacks. - Experience running agents in virtual or sandboxed environments.
Complete your application on jobs.ashbyhq.com. The employer’s form will show what is required.
Already applied? Track this application
Hi, I'm Akbay, Head of Engineering at Reflow.
Most engineering roles at this stage ask you to execute. Pick up tickets, ship features, move to the next one. That's not what we're looking for.
Reflow is building something that requires you to care about the outcome, not just the output. We track how work actually happens inside organizations and turn that into intelligence. The data is messy, the product surface is wide, and the problems are greenfield. You need to want to dig into that, not just be okay with it.
I'm looking for someone who gets energy from complexity, moves fast without cutting corners on the things that matter, and wants to see their fingerprints on a product that real teams depend on. Someone who can build and lead, in that order.
If that sounds like the kind of work you want to be doing, read on.
Akbay
Most organizations have no reliable way to understand how work actually gets done. They track outputs, not workflows. They measure activity, not intelligence. The result is a massive blind spot at the operational layer of every team.
Reflow closes that gap. We're a workforce and workflow intelligence platform that sees how work flows through an organization and turns that into insight and automation. Our automation agents take the workflows we observe and run them for our customers, across browser and desktop apps, with the company-specific context that general-purpose agents don't have. We're early, growing fast, and automation is where we're putting a lot of our weight next.
Month 1: Understand the system and fix what matters most. Learn how a recorded workflow becomes an automation, how a run is executed, and how it recovers. Find the parts that hold the system back the most, fix them, and get the improvements shipped to clients, not just merged. By the end of the month customers should notice the difference.
Month 2: Make the whole system better. Move from the most critical parts to all of them: workflow creation, execution, recovery, evals, observability and cost. Get to a point where we are confident in every update before it ships, and where we can bring on more clients without any drop in quality, speed or cost per run.
Month 3: Drive the project. Take the lead on where automation goes next. Bring your own ideas, sharpen the vision, and keep a close eye on the market: what the model vendors and other agent products ship, what customers ask for, and where Reflow should go to stay ahead. Turn that into a roadmap and steer the project by it.
You own the AI automation project end to end: architecture, reliability, observability, cost, and the roadmap of what it can do next. You stay fully hands-on, this is a building role. Engineers who contribute to automation work with you on it, and you set the technical bar they work to. Beyond automation, you help shape the other AI pipelines in the product, such as our AI companion and the systems that mine our activity data for insight.
Example projects on the horizon:
Build an eval and benchmarking setup that covers our real customer workflows, with regression checks that run on every change.
Push run success rates up across changing UIs, sessions and edge cases, with recovery that fixes the run instead of failing it.
Bring cost per run down with model routing, caching and provider choice, without losing accuracy.
Build end-to-end tracing and observability, so any failed run can be replayed and diagnosed in minutes.
Turn live runs and human feedback into reviewed improvements, so automations get better from use, not just from engineering time.
Shorten the path from a recorded workflow to a working automation, with fewer iterations and no engineer in the loop.
Some engineers think owning a system means being the one who gets assigned its tickets. That's not what we need.
Owning automation means you know how it's doing today without anyone telling you. You know the success rate, the cost per run, the top failure reasons and which customers they hit. When a run fails, you find out why before a customer asks. When a model or prompt change looks good in a demo, you're the one who asks for the benchmark. You decide what gets built next, and you're accountable for whether it worked.
High trust, high speed. We don't manage output through meetings or status updates. You'll know what matters, and you'll have the autonomy to go build it.
Measured, not vibes. An AI change that wasn't measured didn't happen. We ship against baselines, eval sets and production metrics, not against a demo that looked right once.
Cost is part of the design. Inference spend scales with usage. We treat cost per run as an engineering metric, next to accuracy and latency.
AI is part of the craft. We use AI tooling actively across engineering. We expect you to have opinions about what works, for agents in production and for your own workflow.
Direct collaboration. You'll work closely with backend, data and the client app team, and directly with the founders on roadmap and architecture. No long handoff chains.
This role is for you if:
You have built and run multi-step agents in production, with real users, real failure modes and real cost.
You were a strong backend engineer before you worked with LLMs, and you still think like one: retries, idempotency, queues, failure handling, monitoring.
You can answer "how do you know it got better?" with numbers, and you have built the setup that produces them.
You care about cost per run as much as accuracy, and you've made tradeoffs between models, providers and architectures to get both.
You want to own a system end to end and be judged by whether it works in production.
You're energized by an ambiguous, fast-moving space where the playbook is still being written.
You want direct access to the founders and real influence over where the product goes.
This role is probably not for you if:
Your agent experience is mostly RAG chatbots, low-code workflows, or prototypes that never left pilot scale.
You ship prompt changes because they looked better on a few examples.
You think of observability and evals as something to add once things settle down.
You prefer research to shipping, or want a clear spec before you start.
You need someone else to set up infrastructure, deploys or monitoring for you.
You're looking for a pure people management role.
Required
8+ years of backend engineering, with production services and pipelines you designed and owned: APIs, workers, queues, real load, on-call.
4+ years of technical leadership, owning the direction of a system and guiding the engineers who work on it.
Recent, hands-on experience building agentic systems in production: multi-step agents, tool use, MCP, agent loops with recovery and guardrails.
Built or owned eval, benchmarking and tracing for LLM systems: golden sets, regression suites, LangSmith, Langfuse or your own.
Experience managing LLM cost and latency in production: model routing, caching, provider choice.
Strong in TypeScript and Python. You'll work in both.
Architecture judgment: you can take a system that was built fast and set the direction for its next stage.
Strong written communication, we're distributed and context needs to travel well in writing.
Open to working across timezones with a globally distributed team.
Nice to haves
Experience with browser or desktop automation: Playwright, Puppeteer, DOM-based actions, computer-use agents.
Experience with model gateways and providers such as OpenRouter or Bedrock.
Familiarity with Django and ClickHouse, or similar stacks.
Experience running agents in virtual or sandboxed environments.
Previous experience at an early-stage or fully remote company.
Full ownership of the system at the center of where Reflow is going next.
Work directly with the founding team, your architecture decisions will be visible and they'll matter.
Fully remote, work from wherever you do your best work.
A genuinely hard problem: making AI agents reliable, measurable and affordable on real company work.
We offer competitive pay based on the market and where you’re located. The salary ranges in our job postings are intentionally wide because they need to cover both U.S. and international candidates. Our final offer will depend on things like your experience, skill set, and location.
New AI Engineer postings at the employers we follow are up 340% over the last 6 weeks compared with the 6 before.
Each bar is one week of new postings. The darkest bars are the last 6 weeks.
Counted from the hiring pages we’ve followed for at least five weeks, so employers we add don’t look like growth. Updated weekly.
We file this job under AI Engineer, by the work the posting describes.
AI engineers build products on top of AI models that already exist, most often large language models: they connect a model to a company's data and tools and ship the assistants, agents and search features people use.
Postings ask for strong Python or TypeScript, experience with model APIs, retrieval and agent frameworks, and ways to test whether the model's answers are good enough.
An ML engineer trains and serves models; an AI engineer usually starts from a model someone else trained and builds the product around it, though titles often mix the two.
Selected passages from the saved posting. Check the full description for conditions and exceptions.
No pay amount identified in the saved description.
California, United States
- Experience running agents in virtual or sandboxed environments. - Previous experience at an early-stage or fully remote company. What Reflow gives you
- Work directly with the founding team, your architecture decisions will be visible and they'll matter. - Fully remote, work from wherever you do your best work. - A genuinely hard problem: making AI agents reliable, measurable and affordable on real company work.
No clear work-authorization passage found. Eligibility is unconfirmed.