About Furwise
Furwise gives every cat owner help that knows their cat. Owners keep daily care, medications, vet visits and lab results in our app, and an AI agent that knows that history helps them understand what’s going on and when to see a vet. The iOS app is used by cat owners across Asia, Europe and North America, the Android app is in internal testing, and our research-based cat health guides are published in English, Traditional Chinese and Japanese.
About the role
The foundation is already in place: an agent that reasons over each cat’s records, tools that read photos and documents, and a research library in three languages. You’ll own Furwise’s AI from there, and take it where general-purpose models run out: models built for cats, and signals from long-term records.
Cats can’t tell us when something is wrong, and the signs are often small: a slow change in weight, a little more water, a different look in a photo. Your job is to build models that catch those changes early.
Our data is early. Part of the job is deciding what to collect and how to label it, with owners’ consent.
What you’ll work on
- The Furwise agent. Own the agent owners talk to every day: how it reasons over each cat’s records, draws on our research library, and brings your models and health signals to the owner in plain words.
- Early health signals. Find patterns in long-term records, such as slow changes in weight, water intake or litter habits, that may be worth raising with a vet, and flag them early enough to matter.
- Models built for cats. Build and fine-tune models for what owners ask about most, such as pain, mood and stool from a photo, and measure them against the general model we use today.
- Clinical validation. Work with veterinarians to label cases and check what our features say, so we know where they’re reliable and where they should send owners to a vet. We’re building a veterinary review panel, and you’ll help shape it.
- Consent first. Any use of owner data for model development will be opt-in and covered by our privacy policy before it starts. Early work can draw on public datasets and cases labeled with veterinarians.
Your first 90 days
- Learn the current features and what data we have, or can collect with consent, and propose which signal or model to build first.
- Ship a first version, or reach a clear go/no-go on it, measured against a baseline.
- Set up the evaluation and labeling process every model after it will use.
What we’re looking for
- You’ve trained or fine-tuned models on real-world data and put them into production, writing strong Python.
- Solid statistics and evaluation. You know when a result is real.
- Experience with LLM agents, plus time series, health data or computer vision.
- You’ve worked with small or noisy labeled data, using methods such as transfer learning or active learning.
- Fluent English, and care for privacy when you handle health data.
Nice to have
- Experience in veterinary or human health, or working with clinicians.
- Mandarin.
- Published research, or experience running a validation study.
- You have lived with a cat or a dog.
You don’t need to match every item. If most of this describes you, we’d like to hear from you.
This might not be for you if
- You want a large team with a data platform and labelers in place. You’ll build much of that yourself, with the team.
- You’d rather ship fast than validate. When a model says something about a cat’s health, we need to know how often it’s right before it ships.