Four jobs, one agent
Emoti does the legwork between a brief and a decision: finding people, writing to them, running the first screen and keeping the notes. Each piece below is the real product, working on sample data.
Sourcing from a sentence
Write who you need the way you would tell a colleague. Emoti reads it into criteria, checks every profile against each one, and shows its evidence.
More about sourcingEmoti read 6 criteria. Tap one to switch it between must-have and nice-to-have.
At Northstar Search: trains and ships demand-forecasting models. Meets 5 of 6 criteria. Partly meets Time-series forecasting.
Outreach you sign off on
Sequences that open with a line from each candidate's own work, drafted by Emoti and sent only after you approve them.
More about outreachHi Mei Lin,
Your evaluation harness at Gradient Harbor is the kind of work our forecasting team protects time for.
I'm Jordan, and I recruit for Tessellate's forecasting team in Singapore. We predict freight demand a few weeks out…
First screens, any hour
Candidates take a short voice interview when it suits them. The same questions for everyone, scored against your rubric, reviewed by a person.
More about voice screensBuilt Shiplane's first volume forecast. Sales now plans capacity from it, though she couldn't yet say by how much it helped.
Strong answer looks like: names the users, the decision the model changed and a measured result.
Ratings are suggestions. A recruiter listens and decides who moves on.
Notes that point to the moment
Emoti writes interview notes in your template and links every point to where it was said, so the debrief starts from evidence, not memory.
More about interview notesHere's what was said that's most relevant:
Mei Lin Tan [2:15]: "Sure. We forecast demand per lane — origin, destination and equipment type — for the next six weeks. The forecasts feed pricing and capacity planning. I own the modelling side: features, training, evaluation, and the handoff to the serving team. There are about three hundred thousand lanes, but most of the volume sits in maybe twelve thousand of them."
Mei Lin Tan [9:38]: "Honestly, the offline number wasn't enough, and I learned that the hard way. Our first shadow launch looked great offline, but the pricing team found it overreacted to holiday weeks. So we built an evaluation harness that slices error by lane volume, by season, and by the decisions it drives — like how often pricing would have changed a quote. Now every model change ships with that report, and the pricing lead signs off on it."