applied researcher in scientific AI

Prime Intellect, Inc.; United States; Remote; USD 150,000 - 300,000; Full time

Benefits offered by this job

Equity incentives

Visa sponsorship and relocation

Professional development budget

Team off-sites and conferences

Job summary

Prime Intellect, Inc. ищет специалиста по исследованиям в области обучения с подкреплением и пост‑обучения для разработки долгосрочных окружений, обзора научных баз и методик верификации.

Вы будете соединять экспериментальные циклы, данные и требования партнёров в обучающие задачи и сигналы вознаграждения, работая в международной команде. Компания предлагает гибкую локацию и поддержку визы, а также бюджет на профессиональное развитие и участие в конференциях.

Qualifications

• Глубокие знания в RL, агентных системах и пост-обучении
• Опыт создания экспериментальных циклов и воспроизводимых исследований
• Умение строить репозитории кода и экспериментов для совместной работы
• Опыт публикаций/opensource является плюсом

Responsibilities

• Разрабатывать долгосрочные окружения для агентов и их лабораторных исследований
• Преобразовывать данные и требования партнёров в обучающие задачи и сигналы вознаграждения
• Разрабатывать проверяющие методы на корректность и воспроизводимость
• Соединять проверки, симуляции и модели с отзывной обратной связью от экспертов
• Работать с заказчиками и исследовательскими партнёрами на перспективной научной повестке
• Обеспечивать воспроизводимость кода, данных и конфигураций экспериментов

Skills

RL research

Python programming

Experiment design

Multi-agent systems

Job description

Описание

Prime Intellect is building an open superintelligence stack that makes infrastructure used by frontier AI labs available to AI teams. Its Lab platform unifies compute, environments, evaluations, secure sandboxes, high-performance training, and deployment for post-training at frontier scale. The company also develops open‑source frontier AI models and a full‑stack platform for training them.

Задачи

• Build long‑horizon environments where agents search literature, query scientific databases, run code and domain tools, and choose experiments within a fixed compute or experimental budget
• Turn research partners’ workflows, datasets, and outcome criteria into trainable tasks, evaluations, and reward signals
• Design verifiers for correctness, novelty, and reproducibility; test whether agents use underlying evidence and check for label leakage, lookup shortcuts, and unsupported conclusions
• Connect programmatic checks, simulations, domain models, and expert feedback to training; account for noisy or delayed results, calibrate confidence estimates, and define when experiments need human review
• Develop RL and multi‑agent methods for proposing, testing, and critiquing hypotheses; compare them with strong single‑agent baselines at equal compute and evaluate them on held‑out problems
• Keep code, data, logs, and experiment configurations reproducible; rerun promising results, measure noise, use ablations to separate model improvements from changes to data, harness, or inference setup, and release reusable environments, evaluations, and research code
• Work directly with customers, users, or research partners while maintaining a substantive research agenda
• Build experimental loops for scientific tools and datasets, similar to Prime’s autonomous‑research study of multi‑day nanoGPT optimization tasks

Требования

• Depth in RL, agents, or post‑training systems, with the ability to acquire relevant scientific context
• Hands‑on experience building and debugging model‑training, post‑training, agent, evaluation, or closely adjacent experimental loops
• Strong experimental judgment, including meaningful baselines, careful trace inspection, clear failure attribution, and willingness to change direction when evidence changes
• Strong Python skills and ability to build artifacts that other researchers and engineers can run, inspect, and extend
• Comfort owning ambiguous problems and keeping partners and teammates informed while resolving them
• Publications, open‑source systems, benchmarks, or shipped products can demonstrate exceptional work; candidates are encouraged to show something difficult they built, what failed, and how evidence changed their approach
• Будет плюсом: prior work in autonomous research, ML for science, scientific computing, or reproducible computational research

Условия

• Cash compensation range of $150-300k + equity incentives
• Flexible work locations: San Francisco, New York, London, Berlin, or remote
• Visa sponsorship and relocation support
• Professional development budget
• Team off‑sites and conference attendance