Job Overview
Location
Remote
Job Type
Full-time
Category
Software Engineering
Date Posted
January 1, 2026
Full Job Description
đź“‹ Description
- • Own the future of AI-driven software development. As a Member of Technical Staff on the Agent R&D team, you will be the bridge between cutting-edge academic research and production-grade systems that thousands of engineers rely on every day.
- • Translate breakthrough papers into working code. You will continuously scan arXiv, top-tier conferences (ICML, NeurIPS, ICLR, ICSE, FSE), and open-source repositories, then distill the most promising ideas into the OpenHands framework. Your implementations will be the difference between a cool lab demo and a feature that saves real teams hours of debugging.
- • Invent the next generation of coding agents. Beyond porting existing work, you will design and run original experiments that push the frontier of agent reasoning, tool use, and long-horizon planning. Expect to author internal white-papers, file patents, and publish at conferences—your insights will shape the roadmap for the entire open-source ecosystem.
- • Build bullet-proof evaluation harnesses. Quality is non-negotiable when an agent touches production code. You will craft deterministic test suites, probabilistic benchmarks, and human-in-the-loop protocols that measure not just pass@k rates but also security, latency, and maintainability. Your dashboards will be the single source of truth for every release decision.
- • Architect scalable serving infrastructure. From cold-start container pools to streaming inference endpoints, you will design systems that spin up thousands of agents per minute while keeping GPU costs in check. You’ll work with Kubernetes, Ray, Triton, and custom CUDA kernels to shave milliseconds off every token.
- • Champion open-source collaboration. You’ll review pull requests, triage GitHub issues, and host community design sessions. Your documentation and example notebooks will turn casual users into power-contributors, multiplying the impact of every line you write.
- • Influence enterprise adoption. Fortune-500 engineering leaders will look to you for guidance on deploying agents inside regulated environments. You will distill complex research into crisp architecture proposals, security playbooks, and ROI models that unlock six-figure contracts.
- • Shape a culture of kindness and candor. We debate ideas brutally and support people unconditionally. You will mentor junior researchers, pair-program with senior engineers, and give lightning talks that turn “impossible” into “shipped.”
🎯 Requirements
- • Demonstrated research or engineering track record in LLM-based agents, automated code generation, or AI-assisted software engineering (papers, patents, or major open-source contributions)
- • Hands-on experience designing, training, and evaluating agentic systems—bonus for harnesses like SWE-bench, HumanEval, or custom enterprise benchmarks
- • Proficiency in Python and at least one systems language (Go, Rust, or C++); familiarity with CUDA, Triton, or distributed training frameworks a plus
- • Deep understanding of transformer architectures, fine-tuning techniques (LoRA, RLHF, PPO), and scalable serving (Kubernetes, Ray, vLLM, Triton)
- • Comfort working in a fully remote, async-first environment with core collaboration hours 12-4 pm ET (4-8 pm UTC)
🏖️ Benefits
- • Competitive compensation: $150k–$220k base salary plus meaningful equity and annual performance bonus
- • Fully remote with flexible hours and a generous home-office stipend; travel budget for conferences and team off-sites
- • Comprehensive U.S. health, vision, and dental insurance—100% premium coverage for you and dependents
- • Unmetered PTO plus all federal holidays; we actively encourage at least four weeks off per year
- • A learning-first culture: $3k annual education budget, internal paper clubs, and mentorship from founding researchers
Skills & Technologies
R
Senior
Remote
About OpenHands Inc.
OpenHands builds an open-source platform that lets AI agents control a computer like a human—browsing, coding, and running commands in a sandbox. Developers use it to automate software tasks, reproduce bugs, and benchmark agent performance. The project is MIT-licensed, community-driven, and spun out of research at Carnegie Mellon and UC Berkeley.
Similar Opportunities
Faith Technologies, Inc.
Menasha-OMC
Full-time
Expires Mar 4, 2026
Go
Onsite
Degree Required
1 month ago