Research Scientist

Description

The company built a new way to run user research using AI-moderated interviews.

Customers include Hubspot, Microsoft, Uber and Nestle, who use it to gather deep qualitative insight about their users quickly and at large scale.

The market for research services is worth more than $140B and is still growing. Established providers are slow, outdated and expensive, which the company sees as its opening.

Top Silicon Valley investors back the company. In December 2025, less than six months after its Series A, it closed a $30M Series B led by Radical Ventures, with M12 (Microsoft's venture arm), 8VC, Y Combinator and Adverb Ventures participating.

The business grew 8x in the past year as enterprises across many industries adopted AI-moderated research as their default way to understand people.

The team is small and works with some of the largest enterprises in the world.

The company builds digital twins: AI systems that represent specific real people and answer research questions as those people would. The Digital Twins Team runs an ongoing program measuring how closely these systems match human responses and behavior across several levels of analysis.

You will work alongside the Head Scientist to choose topics, design studies, set measures, interpret findings and write them up for internal decisions and, where suitable, academic publication. Your training should cover computational social science, data science, survey methodology, human-computation interaction and experiment design.

Responsibilities

• Run studies that compare system outputs with human responses, using qualitative and quantitative measures.
• Create and refine performance metrics, and justify each one.
• Adapt advanced statistical methods from survey research and large-scale social data analysis to test and raise the accuracy of sample- and population-level simulations.
• Plan evaluations and experiments with analysis plans fixed in advance.
• Lead human rating studies: train raters, write coding rubrics and measure inter-rater agreement.
• Plan data collection from third-party sources and from people directly, including longitudinal and repeated-measures designs.
• Shape the research roadmap for new modeling approaches, including tests of fine-tuned models and model calibration.
• Write analysis specifications detailed enough for an engineer to build.
• Co-author papers, preprints and presentations.

Requirements

• Completed or nearly completed PhD in computational social science, communication, HCI, sociology, political science, psychology, statistics or a closely related field.
• Peer-reviewed publications where you were the lead author.
• Strong command of graduate-level statistics used in social science, including the assumptions behind each method.
• Experience analyzing large telemetry and behavioral datasets and judging whether conclusions are valid.
• Understanding of machine learning across classical, nonparametric and LLM methods, including how transformer language models work, what fine-tuning changes and how major model families differ.
• Working knowledge of computing basics: file systems, software libraries, databases, SQL and tabular data.
• Able to work where most code is written with agentic coding tools, and to read and verify that code.
• Preferred: experience fine-tuning open-weight language models on survey, experimental or behavioral data.
• Preferred: experience building or testing calibrated confidence or error-prediction models, including setting decision thresholds from human judgments.
• Preferred: experience measuring similarity of open-ended text, such as with embedding-based methods.
• Preferred: experience designing or running panel-based data collection with human participants.
• Preferred: experience presenting research at academic or industry conferences.
• Preferred: prior work on generative agents, digital twins or simulation of human respondents.
• Preferred: familiarity with the startup ecosystem.
• Preferred: familiarity with qualitative interviews and ethnographic methods.
• Preferred: familiarity with computational linguistics.

Conditions

Reports directly to the Head Scientist.