Research Engineering Internship

Job description

Twin – Research Engineering Internship **Core Platform · Paris / Hybrid / Remote (Europe)** About Twin Twin builds autonomous agents that are reliable and scalable enough for companies to actually delegate work to them. We engineered the first commercial agent deployed to 500k SMBs — **Invoice Operator** — achieving 95%+ accuracy at a fraction of the human cost. Over the coming months, we expect this first agent alone to automate millions of man-hours. To achieve this, we've built, among other things: - A Rust-based browser infrastructure, maintaining low latency at high scale and concurrency - A password manager enabling secure and automated agent authentication, including solving 2FA - A graph-based agent framework, making complex agents robust through modularity and decoupling - A self-correcting infrastructure, allowing agents to learn from mistakes and continuously improve as they encounter new challenges Founded in 2024, we've raised €12M from LocalGlobe and the founders of companies like Hugging Face, Datadog, and Alan. Our ambition: launch and distribute the next generation of high-impact agents, and become the trusted layer where autonomous work runs. About the Core Platform team The Core Platform team builds the foundational infrastructure powering Twin's high-performance, Rust-based agent engine. This engine manages browser automation, communication between agents and humans, and the evaluation and testing pipelines that keep our agents best-in-class. The team's objective is to continuously improve **speed, cost, and reliability at scale** — and lay the groundwork for training our own advanced models that will **outperform the current state of the art**. About the position This is **first and foremost a software engineering role** — but one with a clear long-term target: the systems you build here will directly enable us to train **our own world-class browsing agents**, surpassing current industry benchmarks in both performance and reliability. In the short term, you'll work on the engineering foundations: better data pipelines, smarter evaluation frameworks, and reliability tooling. In the long term, that work will fuel **the datasets, metrics, and insights** that make our models the best in the world. Main challenges - Build sophisticated **Rust-based data filtering & evaluation systems** to identify the highest-quality agent interactions for training - Develop **large-scale pipelines** to collect, process, and analyze browsing behavior in real time - Create **evaluation frameworks** to measure and compare agent decision quality — and feed those insights back into model training - Solve complex reliability and performance issues with elegant engineering approaches that translate into better datasets and smarter agents Requirements Must-have - Strong software engineering skills and experience shipping reliable code - Creative problem-solving mindset — knowing when engineering beats brute-force model training - Comfortable in lower-level languages (Rust experience a plus) - Interest in AI agents and belief that **great models start with great systems** - Ability to work fast and autonomously - Based in Paris or willing to spend your first month+ in our office (remote/hybrid possible after onboarding) Nice to have - Experience with Rust (any level) - Previous production experience in any technology stack - Understanding of ML concepts (for future model training context) - Experience in data processing, evaluation, or testing infrastructure **No production experience yet?** Show us you're exceptional through personal projects, open-source contributions, competitive programming, or other evidence of your ability to solve complex problems with speed and elegance.