Founding AI Engineer

Why join eLLMo
Brands spend millions on ads before they know whether the page behind the ad will convert. eLLMo fixes that. We run panels of synthetic buyers, each with a distinct personality profile, through a brand's real product page, ad, or email, and return a structured prediction of what will convert and what will not, before the budget is spent. We also measure how a brand shows up when shoppers ask AI assistants what to buy.
Our customers are $20M to $150M DTC and ecommerce brands, and the work ships to real marketing teams every week. You will see what moves their numbers.
We are a small team with two technical founders. The CEO ran strategy analytics at Meta and Palo Alto Networks. The CTO led AI engineering at Gusto. You will work directly with both, and the product is the thing you build.
The role
You will own the intelligence layer of the product: the agents that browse a page like a buyer, the personas that drive them, and the evaluation that tells us whether the output can be trusted. You will also ship the product surface customers touch, from pipeline to dashboard.
There is room to lean toward backend and agent systems, toward the Next.js product, or somewhere between. The one non-negotiable is that you own outcomes, not tickets.
What you'll own
Multi-agent simulation. Design and ship the pipelines that run a full buyer panel in parallel through a page, ad, or email. State, checkpointing, retries, and partial-failure behavior are yours.
Score trust. The same model on the same page should not score 72 and then 61. Build the calibration, regression, and eval harness that makes our scores something a CMO can act on.
Persona quality. Choose, score, and calibrate the personas behind each panel, and build the loop that tells us when they drift.
Model strategy. Route across frontier models by cost and quality, with the observability to make it a decision instead of a guess.
Product, end to end. Run experiments, read the results, and ship the next version. Design to deploy, then iterate on what customers actually do.
What we're looking for
You have shipped LLM or agent systems to production and customers depend on them today. A repo, a PR, or a write-up you can walk line by line.
Experience with production engineering, Python-first, with TypeScript/Next.js fluency a strong plus.
Ownership. You have been the person the customer called when it broke, and you fixed it.
Clear thinking about unknowns. Calibration and score stability are open problems. We want “I don't know” and “here is how I would find out” in the same sentence.
Bonus: show us something you built that you are proud of. Eval harness, observability, forward-deployed, or DTC and paid-media experience all help.
How we work
Schedule: 9am to 6pm, Monday through Saturday. We are in a sprint and we move fast, and we are saying so up front so you can decide if it fits your life.
Location: San Francisco, in person at least 3 days a week. Fast feedback and shared context beat async for us. Remote works for the right candidate.
How we hire
15-minute screen with the CTO.
4-hour technical take-home and meeting with the CTO: walk your own code, pair on a small extension, discuss agent and eval design.
15-minute conversation with the CEO.
Paid 3-day trial on real backlog work.
Application to offer in 1 week.

Work arrangement
Yes