Product Engineer at AI-native ecommerce startup

⚡ About Tellis
Tellis is an AI-native product data enrichment platform for ecommerce brands.
Large product catalogues leak revenue. Turning a spec into a live product page takes weeks. Most teams only have bandwidth to polish their top 20% of SKUs, and thin or messy data across the rest costs brands millions a year in lost sales, returns, and compliance issues. With the rise of agentic commerce, this problem is now urgent: if merchants' product data isn't AI-ready, they become invisible to shopping agents like ChatGPT Shopping and Google AI Mode.
Tellis is a team of AI agents that sits on top of a merchant's stack, turning siloed shopper signals into enriched, validated product data at scale, pushed straight into the client's CMS or PIM. We're building the AI shopping context layer that makes every product discoverable and conversion-ready for the AI shopping era.
We're a VC-backed, pre-seed startup with paying customers in the UK, founded by ex-Google and second-time founders.
🧑‍💻 The role
You'll be an early engineer on the team that turns messy merchant catalogues into enriched, validated product data.
The job has two halves: working out what a merchant's catalogue needs and where its opportunities are, then doing the engineering that turns that understanding into product — first as customisation for one account, and after that as a feature every account gets. Your clients tell you what to build, based on what you put in front of them.
Every new account brings a different data source, a business need, a different category tree, a different definition of good copy.
What you’d do:
You own retention on one or two merchant accounts.
Your read of their catalogue shapes how they see our opportunities engine, and you're judged on the initiatives they accept and that go on to work.
You decide what a merchant should be able to change without a code change, and how the core has to adapt to absorb it — then you build it: the per-customer policies, prompt overlays and mapping documents that let a merchant be configured rather than coded.
You make the platform absorb each account's difference instead of working around it — configuration when that's enough, a pull request against the product code when it isn't.
Connector customisation is yours, with the core team behind you.
An escalation has three possible answers: a new setting, a roadmap change, or no. Never a branch, a customer-named flag, or a script run against production.
Leaning fully on agentic development is how you'll hit the prototyping speed this needs, but a PR into the core exists to make the product more general across clients, never to fit it to one.
This is the way into owning the integrations and opportunities layer at Tellis. As the book of business grows, the first product engineer on it becomes the person who turns one-off technical asks into a repeatable product engine.
✅ Must have
Extensive ecommerce background: you've worked on catalogues, product data, or consumer brands. You can look at a PDP and say why it converts, and judge whether generated copy is right for a given retailer.
You can find the opportunity in a catalogue using the data a merchant already has — GA4, Search Console, Merchant Center, product feeds, their own sales numbers. Knowing which products are worth fixing is the job; knowing which ones score badly isn't.
A track record of deciding what gets built, not only building it.
You bring clients opportunities they haven't asked for.
Real depth in at least one commerce platform's API and data model: Adobe Commerce, Shopify, BigCommerce, commercetools, or a major PIM. Connector work is yours.
Strong Python, and comfortable in MongoDB — documents, aggregation pipelines, and the mess that accumulates in a production collection. You'll be shipping against the product codebase, not only configuring it.
LLM-native both ways: you've built with LLMs in production (agent workflows, evaluation) and you work agent-first with Claude Code, Codex or equivalent. Our prototyping speed assumes the second; knowing why an agent fails and how to prove it got better is the first.
❇️ Nice to have
Adobe Commerce specifically — it's where our accounts are today
A second commerce platform beyond your primary one
You've owned a settings, admin or configuration surface as a product, and have scars from it
Breadth of catalogue exposure: agency-side across many merchants, or time at a PIM, feed or merchandising vendor
Startup experience, or evidence you've shipped things without being told how
🎁 What you get
Ground-floor role at a funded startup with early revenue and a clear thesis
Direct work with founders and clients
Competitive compensation (depends on experience)
Clear ownership and opportunity to expand responsibilities
Remote work arrangements

Work arrangement
No