Job description

ABOUT LIGHTFIELD Lightfield is reimagining CRM as a world model of a business. By learning from emails, meetings, and customer conversations, we’re building a living understanding of how a company works—so AI can help teams anticipate what comes next, make better decisions, and take action. More than 5,000 companies have used the product since its November launch. We’ve raised more than $70M from Andreessen Horowitz, Coatue, Maverick Ventures, and Greylock to attack the biggest software category in the world. We previously built Tome, which reached more than 25 million users, and our team brings experience from Llama, Instagram, Google, and Salesforce. ABOUT THE ROLE We’re starting Lightfield Research to build a Business World Model through frontier research in simulation, world modeling and self-improvement. We’re hiring applied researchers and research scientists to lead its first projects. You’ll report directly to Ves Stoyanov, co-founder and Chief AI Officer, previously at Facebook AI, where he co-authored RoBERTa and XLM-R. You’ll work alongside AI engineers who take research into production and GTM experts who help define tasks and quality. Researchers at Lightfield own projects from the first question through measurable product impact and publication. You’ll help shape the lab’s research agenda, culture and reputation as it grows. WHAT YOU’LL DO - Lead research projects in evaluation, benchmarks, deal prediction, simulation and self-improvement. - Design empirical evaluations of Lightfield’s AI system using LLM-as-judge, rubric-based grading, offline replay and A/B tests, on both production and offline data. - Build benchmarks for core GTM tasks such as list building, CSV imports, mail intelligence, record updates and outbound. Measure Lightfield against frontier models and competitors. - Build simulated companies with users and streams of email, meetings and knowledge, where the right answers are known. Use them to evaluate the full system end to end. - Develop techniques that let Lightfield improve as it operates, at the world, company and user levels. These include reinforcement learning and, where it pays off, fine-tuning. - Model deal probability and the factors behind it, and ship them as product features. - Improve cost and quality trade-offs across the AI system, including inference cost. - Publish research blog posts, open benchmarks, arXiv reports and papers, and present at conferences and meetups. - Help build a world-class research team by recruiting and mentoring teammates. WHO YOU ARE - You have 7+ years of applied research experience in ML or NLP, in industry or academia. - You have a PhD or MS in Computer Science, Machine Learning or a related field, or equivalent research experience. - You have a track record of research that shipped or was published: papers, open-source models, datasets or benchmarks. - You have deep hands-on experience with LLMs, including evaluation, agent design and retrieval. Post-training experience (SFT, RL, preference optimization) is a plus. - You are rigorous about experimental design and can tell a real improvement from noise. - You write strong Python and are comfortable working in a production codebase (TypeScript, Node.js, PostgreSQL) alongside engineers. - You can drive an ambiguous problem independently and balance ambitious goals with frequent, visible results. - You communicate clearly to technical and non-technical audiences, in writing and in talks. - You are energized by building AI that businesses depend on to do their jobs. NICE TO HAVE - Experience building simulations, synthetic data or RL environments. - Experience with forecasting or predictive modeling on business data. - Experience creating benchmarks adopted by others. BENEFITS & PERKS - Competitive salary - Meaningful early equity - Health insurance (medical, dental, vision) - 3 weeks of PTO - 11 paid company holidays + we enjoy a winter holiday break - 3 months of paid family leave - Wednesdays work from home - Regular team dinners, events, offsites, and retreats - 401k plan - Other perks include: commuter and lunch stipend

Requirements

Must-haves

Researchers at Lightfield own projects from the first question through measurable product impact and publication. You’ll help shape the lab’s research agenda, culture and reputation as it grows.
Lead research projects in evaluation, benchmarks, deal prediction, simulation and self-improvement.
Design empirical evaluations of Lightfield’s AI system using LLM-as-judge, rubric-based grading, offline replay and A/B tests, on both production and offline data.
Build benchmarks for core GTM tasks such as list building, CSV imports, mail intelligence, record updates and outbound. Measure Lightfield against frontier models and competitors.
Build simulated companies with users and streams of email, meetings and knowledge, where the right answers are known. Use them to evaluate the full system end to end.
Develop techniques that let Lightfield improve as it operates, at the world, company and user levels. These include reinforcement learning and, where it pays off, fine-tuning.
Model deal probability and the factors behind it, and ship them as product features.
Improve cost and quality trade-offs across the AI system, including inference cost.
Publish research blog posts, open benchmarks, arXiv reports and papers, and present at conferences and meetups.
Help build a world-class research team by recruiting and mentoring teammates.
You have 7+ years of applied research experience in ML or NLP, in industry or academia.
You have a PhD or MS in Computer Science, Machine Learning or a related field, or equivalent research experience.
You have a track record of research that shipped or was published: papers, open-source models, datasets or benchmarks.
You are rigorous about experimental design and can tell a real improvement from noise.
You write strong Python and are comfortable working in a production codebase (TypeScript, Node.js, PostgreSQL) alongside engineers.
You can drive an ambiguous problem independently and balance ambitious goals with frequent, visible results.
You communicate clearly to technical and non-technical audiences, in writing and in talks.
You are energized by building AI that businesses depend on to do their jobs.

Nice-to-haves

You have deep hands-on experience with LLMs, including evaluation, agent design and retrieval. Post-training experience (SFT, RL, preference optimization) is a plus.
Experience building simulations, synthetic data or RL environments.
Experience with forecasting or predictive modeling on business data.
Experience creating benchmarks adopted by others.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

aillmmlmachine learningpythontypescriptnode.js

Platform: ashby
Effort: quick

Benefits

Benefits detected in the listing.

Health Insurance

Dental

Vision

401k

Pto