Help every family remember.
We're building the infrastructure for generational memory: one place where a family's stories, voices, and history are kept. Come build it with us.
Right now a family's story is scattered.
Across phones. Apps. Albums. Attics.
And every year, a little more of it is gone.
We're building the one place it's kept, for good.
How we work
The mission is the boss.
Every decision starts with one question: does this help capture a story that would otherwise go untold? If not, we don't do it.
Ship today.
Someone's parent is getting older while we talk about it.
Own it end to end.
You take your work from idea to production. Nobody sits between you and the result.
Kind over nice.
A nice person ignores the problem. A kind person tells you about it. We'd rather be kind.
Care about what nobody sees.
Our users won't know how we built this. But if we cut corners, they'll feel it.
Mistakes are investments.
Make them fast and learn from them. The only one we mind is the same one twice.
Results, not hours.
We don't track hours. We track what shipped and what it changed.
Open roles
No CV. Show us something you've made.
One day a small marketing team will run ten thousand experiments a week and never go off brand. We are building that team today.
We are building marketing that acts on its own. It plans a test, builds the ad, the page or the email, runs it, reads the result, and ships what won. Then it does it again, a thousand times, across everything a customer sees.
Ethos is the platform where families keep their stories forever.
Your job is to build the machine. A freelancer gives us a solid landing page or a solid ad. You build the system that takes that start and makes it better on its own, test after test. The system will run on its own, and marketing dollars become more families who remember their stories and their wisdom. We are building a world where nobody is forgotten.
You'll learn to build marketing with LLMs at a level very few people can yet: multi-agent orchestration, eval pipelines, closed-loop optimization, guardrail design.
You will learn a ton here. We all learn a ton every day, and you absolutely have to like that. Years of experience matter less than how fast you learn. A fixed way of doing things is a problem here.
What you need
- Python, with AI. You can write a script that calls an API and reads the result. Yes, writing it with AI counts.
- You know what a p-value is. You have run a real test, in a thesis or at work, and you know when a test is not done yet.
- An ad or a piece of content. You have made one, and spent real money or real hours to see what happened.
- Tallinn, in person.
What we can't teach you
- Taste. You know good from bad, and on brand from off.
- Trust in the numbers. You back every claim with data, even when the data says you were wrong.
- Systems thinking. You see the system loop, not the task.
- Understanding people. You can change a room's mind and understand where the AI does not understand the emotional nature of a human being.
- Hacker mindset. You set up a Minecraft server once, jailbroke a phone, hacked a game or a system, or taught yourself something nobody asked you to.
- How to strive to be extraordinary.
What you get
- Base pay + a solid motivational package based on the growth you drive.
- After a trial period, a real chance to come on full-time.
Skip the résumé. Share with us something you've done.
Humanity forgets almost everything. We're building the system that stops that: an interviewer better than any human, and a family archive built to outlast us all.
Ethos is where families keep their stories forever. One person talks about their life. Over time that grows into a family archive: every voice and every memory, kept across generations.
The role. Second engineer, next to the CTO. You decide what gets built together with the product manager. Then you build it, watch real families use it, and fix what you got wrong.
How we build. With coding agents, several at once. Typing speed is worth nothing here.
What you'll work on
- An interviewer that learns. Every question is a bet. We measure what each one brings back. It learns which question to ask this person right now, when to try something new, and when to go quiet. The reward signal is the hard part: define it a little wrong and it learns to pad.
- Context engineering. What goes into the model before each question decides most of the quality.
- A data architecture that outlives the models. The archive is the product. The models are a dependency.
- A voice interviewer. Most of our storytellers are past 70. It does not cut them off.
- A network of family knowledge. Stories linked across a family, and the gaps worth asking about.
How we build
- Architecture first. You know what good software architecture looks like. Agents produce what the structure allows.
- Review and nudge. You read what an agent wrote, spot the overcomplexity, and steer it toward something elegant and robust.
- Delegate between agent teams. You split a problem so several agents run it in parallel, and you know which pieces never to hand off.
- Guidance that lasts. You write the instructions and guardrails that make an agent get it right the first time, and you keep them alive as the code changes.
Who this is for
- You like difficult problems. The kind nobody has solved yet, with no spec and no answer key. You stay on them until they give.
- Reward design. You have designed a reward function, shipped it, and watched it get gamed. You know the gap between what's easy to measure and what you actually want.
- Systems that learn from their own choices. A recommender, a policy, a bot that plays a game. You know why the greedy move loses.
- Context engineering. You have built memory for an LLM system: what to load, what to keep, what to drop.
- You're able to think in graphs. Themes, threads, people, places. You have searched one for the best next move, not just the nearest.
- Data architecture. You have designed a data model that had to last longer than the code around it.
- Empathy for someone nothing like you. An 82-year-old widow in Utah, mid-sentence, deciding whether to trust a machine with the hardest year of her life. You know what she needs to hear next. That counts as much as the math.
- Agentic coding, every day. You can show us the work, and the guidance behind it.
- Python and TypeScript. One strong, the other you pick up fast.
- Tallinn, in person.
Who this is not for
- You need a spec before you start.
- You've never shipped anything to a stranger.
- You want to be told what to build.
What you get
- Base pay plus a package that grows with what you ship.
- Direct line to the CTO and the founders. Second engineer in, so you shape how the team builds.
- Trial period, then a real full-time offer.
No résumé. Send one thing you built that learned, or one thing you built with agents, and what it changed.
We don't have a title for this one.
Ethos is building the platform where families keep their stories forever. Early stage, big mission, entering the US market right now.
We don't know what you'd do yet. We do know the feeling when the right person walks in.
You might be a marketer, a storyteller, an engineer, or very good at something we haven't thought of.
If something about Ethos gives you a feeling, don't overthink it. Write to us. Tell us what you'd bring.
FAQ
No. Send us something you've made and a few lines on why this matters to you.
Try the product and read our story. The best first conversations happen when you already get why this matters.
Salt Lake City and Tallinn, in person.
Small and early. You own real work and answer for real results. Feedback is direct. We move fast because the mission needs us to.
Reach out anyway. If you're great and this mission is yours, we'll figure it out together.