01

Careers

Laplace Research is a small team in San Francisco and New York working on models whose confidence can be taken at face value. We care about careful measurement, honest claims, and releases other people can check.

We hire rarely and review applications on a rolling basis. If one of the roles below fits, we would like to hear from you.

02

Open roles

2 positions

Research engineering

Member of Technical Staff, Forecasting Data

Location
San Francisco or New York
Type
Full-time

Build the corpus that calibrated models learn from: resolved questions from markets, history, and expert judgment, with outcomes that are well specified and consistently resolved.

What you will do

  • Design pipelines that collect, deduplicate, and resolve forecasting questions across venues.
  • Write resolution criteria and checks that keep outcomes unambiguous.
  • Guard against leakage between training data and the future it is meant to predict.
  • Prepare datasets, datasheets, and splits for public release.

What we look for

  • Strong Python and experience building data systems that others rely on.
  • Care about provenance, versioning, and reproducible pipelines.
  • Comfort reading probability and evaluation code.
  • Interest in forecasting, prediction markets, or measurement.

Research

Member of Technical Staff, Calibration Research

Location
San Francisco or New York
Type
Full-time

Train and evaluate models whose probabilities mean what they say, and study the mechanisms that make their confidence trustworthy.

What you will do

  • Run post-training experiments, from supervised fine-tuning to reinforcement learning, aimed at calibration.
  • Build evaluations like ConvictionBench that separate evidence from pressure.
  • Investigate how models represent uncertainty and when that representation fails.
  • Write up results with the evidence attached, including what did not work.

What we look for

  • Hands-on experience training or fine-tuning language models.
  • Fluency with evaluation design, statistics, and uncertainty estimates.
  • A record of careful experiments: papers, open-source work, or thorough write-ups.
  • Interest in calibration, interpretability, or forecasting.

03

How we hire

  1. i

    Application

    Tell us about your work and share links we can read. A founder reviews every application.

  2. ii

    Conversation

    A call about what you have built, what you want to work on, and how we work.

  3. iii

    Technical deep dive

    A focused discussion of a project of yours, or a short problem close to our research.

  4. iv

    Decision

    We aim to reply at each step, whether or not we move forward.

04

Apply

Fields marked with an asterisk are required.

We use this information only to consider your application.