Latest AI JobsBrowse Jobs
Cognition logo

Research Engineer, Post-Training

CognitionSan FranciscoResearch & Development
Apply on company site
Salary
Not listed
Job type
Full-time
Posted
6 Aug 2026
Status
Open · checked 7d ago

About this role

Cognition is currently seeking a Research Engineer, Post-Training to join their team in San Francisco. This full-time role focuses on advancing AI agent capabilities and suits professionals passionate about solving complex machine learning challenges.

Role Overview

The post-training phase serves as the essential link between raw model potential and safe, practical real-world utility. In this position, candidates will refine how artificial intelligence systems learn by constantly improving training workflows, assessment strategies, and alignment techniques. This opportunity requires a unique combination of profound scientific inquiry and practical software execution, operating within a high-velocity environment where research and product development happen simultaneously.

Key Responsibilities

Design and refine training recipes by managing datasets, training phases, and hyperparameters to directly influence model behavior and production performance.
Create robust evaluation frameworks that measure meaningful outcomes and accurately capture real-world capabilities.
Investigate unexpected training outcomes thoroughly to gather insights that inform future problem-solving.
Implement and advance alignment methodologies like RLHF, RLAIF, and constitutional frameworks to guide agent reasoning and long-horizon tasks.
Analyze performance scaling relative to data and compute resources, inventing new strategies when standard methods reach their limits.

Requirements

Proven history of improving machine learning systems through post-training, alignment, preference modeling, reward learning, or related techniques.
Solid foundation in machine learning theory, statistics, and probability to accurately interpret experimental findings.
Demonstrated history of original contributions through top-tier publications, open-source achievements, or relevant industry accomplishments.
Background handling large-scale distributed training environments and resolving associated technical issues.
Systems-oriented mindset capable of connecting training pipelines, data management, and evaluations.
Ability to thrive in ambiguous, fast-paced settings with minimal operational bureaucracy.

Skills and Tools

Machine Learning Theory and Practice
Post-Training and Alignment (RLHF, RLAIF)
Preference Modeling and Reward Learning
Large-Scale Distributed Training
Data Analysis and Statistics
Evaluation Design

About the Company

Cognition operates as an applied artificial intelligence lab focused on creating end-to-end software agents, notably recognized as the creators of Devin, the initial AI software engineer. The collective consists of competitive programmers, former founders, and industry veterans from prominent technology organizations. Team members enjoy access to massive compute allocations spanning thousands of GPUs, fostering an environment defined by speed, autonomy, and technical excellence.

How to Apply

Interested candidates should submit their application directly on the official company careers page.

Questions candidates ask

Where is this job located?

This position is located in San Francisco.

What is the primary focus of the Research Engineer role?

The role focuses on post-training, which bridges raw model capability and useful real-world system performance through training recipes, evaluations, and alignment methods.

What kind of company is Cognition?

Cognition is an applied AI lab that builds end-to-end software agents and created Devin, the first AI software engineer.

What qualifications are expected from candidates?

Candidates should have a track record in post-training or alignment, strong ML and statistical fundamentals, original contributions like publications or open-source impact, and experience with large-scale distributed training.

How should I apply for this position?

Candidates can apply directly on the official company careers page.

Ready to apply?

The application is on the official Cognition careers page.

Apply on company site