Job Overview
Location
San Francisco Bay Area
Job Type
Full-time
Category
Data Science
Date Posted
May 3, 2026
Full Job Description
đź“‹ Description
- • Research, Post-Training role at Cognition, an applied AI lab building end-to-end software agents like Devin, the first AI software engineer, and Windsurf, an AI-native IDE. This role is critical in bridging raw model capability with real-world usefulness, safety, and effectiveness by shaping how agents learn through training recipes, evaluations, and alignment methods.
- • Day-to-day responsibilities include developing post-training recipes by iterating on datasets, training stages, and hyperparameters; designing and maintaining evaluation integrity to ensure metrics reflect real-world performance; diagnosing unexpected training results to root causes and applying insights forward; advancing alignment techniques like RLHF, RLAIF, and constitutional methods to shape agent reasoning and collaboration; and exploring scaling behaviors with data and compute to develop new methodologies when existing approaches plateau.
- • The team is small, talent-dense, and composed of world-class competitive programmers, former founders, and researchers from leading AI organizations including Scale AI, Palantir, Cursor, and Google DeepMind. Research and product development are tightly integrated, with prototypes rapidly deployed to real users.
- • In this role, you will deepen your expertise in post-training and alignment, contribute original research or engineering advancements, work at frontier scale with massive compute resources, and directly influence the capabilities of AI agents that collaborate with humans on complex, long-horizon tasks.
🎯 Requirements
- • Track record of advancing ML systems through post-training, alignment, RLHF, RLAIF, preference modeling, or reward learning
- • Strong fundamentals in probability, statistics, and ML theory with ability to distinguish real effects from noise or bugs
- • Evidence of original contributions via top-tier publications, open-source impact, or equivalent industry results
- • Experience with large-scale distributed training and associated debugging
- • Systems-level thinking: understanding how training pipelines, data, and evaluation interact beyond model optimization
- • Comfort with ambiguity and fast-moving research environments where priorities shift quickly
🏖️ Benefits
- • Compute is not a constraint: large allocations with training jobs running across thousands of GPUs from day one
- • Small, highly selective team where research and product move together; prototypes reach real deployment quickly
- • Environment rewards speed, autonomy, and technical depth with minimal process overhead
- • Opportunity to work on one of the most competitive and fast-moving problems in AI: end-to-end software agents
- • Equal opportunity employer committed to reasonable accommodations for candidates with disabilities
Skills & Technologies
About Cognition
Cognition is an applied AI lab renowned for creating Devin, the groundbreaking AI software engineer. This innovative platform is designed to automate complex software development tasks, serving enterprise clients and teams seeking to enhance efficiency in their engineering workflows. Devin's capabilities extend to areas like automating .NET framework migrations and serving as a data analyst, as demonstrated by customers like Eight Sleep. The company's credibility is underscored by its talent-dense team, whose founders boast 10 IOI gold medals and prior experience from leading AI firms such as Google DeepMind and Scale AI, positioning Cognition at the forefront of AI-driven software development.
Subscribe to the weekly newsletter for similar remote roles and curated hiring updates.
Newsletter
Weekly remote jobs and featured talent.
No spam. Only curated remote roles and product updates. You can unsubscribe anytime.
