Applied Research Engineer, Agentic Code Generation

Description A bit about us: Tabnine is redefining how software gets built. Trusted by over 1M+ developers, we build AI-first developer experiences powered by state-of-the-art coding agents and code reasoning models. With support for 30+ programming languages and 15+ IDEs, our platform is pushing the limits of LLM-based software engineering - enabling teams to design, write, review, and ship code faster than ever. We’re committed to advancing code-native AI models, multi-agent systems, agent orchestration frameworks, memory, and autonomous dev tooling to empower developers at every step of the software lifecycle. We’re growing fast, and our team is passionate about pushing AI engineering to new heights - solving complex problems in LLM training, inference optimization, reasoning, and agent orchestration at scale. About the Role: As an Applied Research Engineer you’ll work on cutting-edge code-focused LLMs and AI agent systems that power Tabnine’s next-generation developer platform. You’ll be at the center of research, model training, and productionization of intelligent systems that understand software deeply, collaborate with developers, and help automate engineering workflows end-to-end. Your work will immediately impact millions of engineers worldwide. Responsibilities Responsibilities: Push LLM Innovation: Research, design, and fine-tune domain-specific LLMs for code generation, refactoring, debugging, and multi-turn reasoning. Agent-Oriented Development: Build multi-agent coding systems that integrate retrieval-augmented generation (RAG), code execution, testing, and tool use to create autonomous, context-aware coding workflows. Production-Grade AI: Own the training-to-inference pipeline for large code models—optimize inference with quantization, distillation, and caching techniques. Rapid Experimentation: Prototype and validate ideas quickly; leverage reinforcement learning, human feedback, and synthetic data generation to push accuracy and reasoning. Cross-Functional Collaboration: Partner with product, engineering, and design teams to ship AI-powered features that help developers focus on high-impact work. Scale the Platform: Contribute to distributed training, scalable serving systems, and GPU/TPU-efficient architectures for ultra-low-latency developer tools. Requirements Requirements: 2+ years of hands-on experience designing, training, and deploying machine-learning models M.Sc. or higher in Computer Science / Mathematics / Statistics or equivalent from a university, or B.Sc. with strong hands-on ML experience Practical experience with Natural Language Processing (NLP) and LLMs Experience with data acquisition, data cleaning, and data pipelines A passion for building products and helping people, both customers and colleagues All-around team player, fast, self-learning individual Nice to have Nice to have: 3+ years of development experience with a passion for excellence Experience building AI coding assistants, code reasoning models, or dev-focused LLM agents. Familiarity with RAG, function-calling, and tool-using LLMs. Knowledge of model optimizations (quantization, distillation, LoRA, pruning). Startup or product-driven ML experience, especially in high-scale, latency-sensitive environments. Contributions to open-source AI or developer tools.