Post-train, fine-tune, and distil frontier language models using RL, distillation, and reward design in Python. Build production-grade simulation systems that model thousands of individuals in real-time for a Societal World Model at a YC-backed startup.

You’ll work with Python, FastAPI, TypeScript, React, and GCP. Electric Twin builds a behavioral simulation platform that creates “digital twins” of real populations, helping organizations test decisions with 95% accuracy.

You will work with Python, PyTorch or JAX, Docker, CI/CD, and cloud platforms to deploy ML models for protein design. The team builds scalable infrastructure for training, inference, and evaluation across molecular and protein datasets, taking research prototypes into production.

You will work with agentic LLMs, reinforcement learning, Python, PyTorch, JAX, HF, and tools like CleanRL/AutoGen/LangGraph on an agentic code generation tool. The team designs multi-agent and RL approaches, builds research prototypes, creates evaluation suites, and curates datasets from complex technical documents.

EXL Service★3.7 (2k)· London

You will work with Generative AI tools (OpenAI, LangChain, Hugging Face, Vertex AI, Bedrock), LLMs, vector databases (FAISS, Pinecone, ChromaDB), and Python frameworks (Transformers, PyTorch, scikit-learn). As part of EXL’s Data Science Team, you’ll build and deploy GenAI solutions for enterprise use cases such as customer service, document automation, summarization, and knowledge retrieval.

Sonar★5.0 (2)· London

You will work with Python, core ML frameworks, and post-training techniques including GRPO, PPO, RLVR, SFT, PEFT, and preference/safety alignment (Rust or SonarQube flagship languages are a plus). You will join a cross-disciplinary team developing advanced products that post-train models to power enterprise agentic coding practices, ensuring code meets enterprise standards.

You'll work with Python, Go, React, TypeScript, Docker, Kubernetes, APIs, model-serving systems and distributed-inference workloads. Partnering directly with customers, you'll architect, build and deploy production AI/ML systems on Lightning's end-to-end platform, covering inference infrastructure, distributed systems and AI workflows.

You will use Python, core ML frameworks, and post-training techniques (GRPO, PPO, RLVR, SFT, PEFT) to develop enterprise-grade coding agents. This cross-disciplinary team combines Sonar’s static analysis with LLM post-training to help customers build agents generating high-quality code meeting enterprise standards.

You will work with Generative AI, LLMs, OpenAI, LangChain, Hugging Face, PyTorch, TensorFlow, FastAPI, Streamlit, and cloud platforms (AWS/GCP/Azure). You will design and deploy GenAI solutions—including RAG pipelines, fine-tuning, and prompt engineering—for enterprise use cases such as customer service, document automation, and knowledge retrieval on the Data Science Team at a global data and AI company.

Train and fine-tune Vision-Language Models (VLMs) and multimodal foundation models using Python and PyTorch or JAX, implementing research from arXiv into production code and owning the full model lifecycle. The company builds large-scale models for autonomous physical-world capabilities, aiming to protect democracies via technological leadership.

Sonar★5.0 (2)· London

Sonar seeks a senior researcher with an advanced degree and 4+ years of ML industry experience. The role involves developing enterprise-grade coding agents using Python, ML frameworks, and post-training techniques (e.g., RLVR, SFT) for Sonar’s AI code verification platform. You will work within a cross-disciplinary team to translate prototypes into products.

Work with Rust, Python, Java, Kotlin, C++ and related languages, building a next-generation software platform that integrates robotics, sensors and autonomous systems for demanding environments in the security sector.

You'll work with PyTorch, JAX, TensorFlow, NCCL, MPI, Slurm, Kubernetes, and Ray. Join a cross-functional team building core tools for high-volume data processing, RL agent training, and large-scale foundation models to protect democracies. You'll extend deep learning frameworks, scale distributed training, and debug production ML pipelines.

Optimize ML inference for edge accelerators and GPUs using TensorRT, CUDA, Qualcomm QNN, Triton, OpenCL, MLIR, ONNX, Python, and C++. You will run large transformer-based models efficiently on low-power edge devices to enable Wayve’s first driving product.

You will use Python, Trio, PyTorch, TensorFlow, JAX, Kubernetes, Rust, C++, and GPU clusters. The team advances reinforcement learning for large language models, improving Claude’s autonomy, coding, and reasoning capabilities.

Design and develop low-latency, high-throughput trading and market data systems using modern C++ (C++17/20) on Linux, with additional exposure to Python, Rust, Go, Java, or C#. The team handles strategy execution, market data processing, and core infrastructure on the direct trading path.

You'll use Python, machine learning, reinforcement learning, LLMs, and Kubernetes to run experiments like AI Debate and jailbreak evaluations. As part of Anthropic's Alignment Science team, you'll conduct exploratory AI safety research on risks from powerful future systems, covering AI control and alignment stress-testing.

Research Engineer will work with large-scale ML systems, GPUs, Kubernetes, and PyTorch to optimize attention mechanisms, compare Transformer variants, and scale distributed training. They conduct research on model architecture and algorithms for the Pretraining team at Anthropic, a company building beneficial and safe AI systems.

Deliveroo★3.3 (1.5k)· London

Build production infrastructure for Generative AI at Deliveroo and DoorDash, leading design and architecture for real-time GPU serving, high-throughput batch inference, and fine-tuning (SFT/DPO/LoRA) of open-weight LLMs and VLMs. Own the LLM Gateway, Agent Gateway, evals infrastructure, guardrails, and cost attribution, while mentoring engineers and setting technical direction for a small, high-leverage centralized GenAI platform team.

You will work with GenAI tools like OpenAI, LangChain, Hugging Face, Vertex AI, and Bedrock, plus Python, PyTorch, TensorFlow, vector databases (FAISS, Pinecone), and cloud platforms. The Data Science Team builds and deploys LLM-based applications for enterprise use cases such as customer service, document automation, summarization, and knowledge retrieval.

Working with Python, C++, and deep learning frameworks such as TensorFlow, PyTorch, or JAX, the role covers LLMs, deep learning optimization, reinforcement learning, and generative models. The research team applies AI to semiconductor design and optimization, alongside fundamental research supporting internal applications and the broader scientific community.

You’ll implement ML research ideas using Python or C++, alongside deep learning frameworks like TensorFlow, PyTorch or JAX. The team applies AI (including LLMs, reinforcement learning and generative models) to semiconductor design and fundamental research.

You'll design, build, and maintain scalable AI data pipelines, cloud-native backend systems, and distributed services using Python, REST APIs, and microservices with FastAPI, Flask, or Django. You'll productionise Machine Learning, NLP, and Generative AI models, work with SQL databases, AWS/Azure, Docker, and Kubernetes, and mentor juniors. The company specialises in AI and data engineering, transforming AI research into production systems.

You will use Python, pandas, scikit-learn, PyTorch, and Hugging Face to wrangle messy legacy data and build production-ready AI models. The team is building an AI operating system for industries with fragmented, complex data, shipping models for forecasting, optimisation, and recommendation.

Legora★0.0· London

You will use Python, data analysis, surveys, interview protocols, and dashboards to study how lawyers and legal teams build proficiency with AI tools. The AI Transformation & Applied Research team turns this evidence into actionable recommendations for product, customer enablement, and go-to-market at Legora, an AI-native legal workspace.