AI Researcher - Post-Training

Position description

At Sonar, we are seeking an ambitious senior researcher to join our cross-disciplinary team, innovating and developing the next generation of solutions to build enterprise-grade coding agents and models. You will harness Sonar’s deep experience in static analysis, and combine it with your experience and leading techniques in large language model post-training. If you are interested in being hands-on with state-of-the-art research, building practical solutions that deliver high-impact for customers, and working within a team of innovative researchers and engineers, this role is for you.

What you will do

• Outcome Driven Development: Work in a team developing and implementing advanced products that enable customers to post-train models to power their agentic coding practices. These agents need to generate high-quality code that meets their enterprise standards and software development best practices.

• Translate Prototypes to Products: Collaborate closely with researchers, research engineers, MLOps and engineers within the team to design hypotheses and experiments, iterate proofs-of-concept quickly and develop successful prototypes into cutting-edge products.

• Subject Matter Expert: You will contribute and discuss ideas within our cross-disciplinary team, driving towards the next generation of coding model post-training for enterprises.

• Spearhead Research & Innovation: Stay up-to-date with the latest LLM and agentic developments; you are driven by learning and teaching others. You will need to explain complex technical details and concepts to both technical and non-technical audiences.

Experience and qualifications

The ideal candidate will have:

• An advanced academic background (Master’s or PhD) in Computer Science, Machine Learning, or a related quantitative field.

• 4+ years industry experience in machine learning, with a solid understanding of modern software engineering practices and tools.

• Fluency with Python including core ML frameworks, experience with Rust or any of SonarQube’s flagship languages (C#, C++, JS/TS, Java) is a plus.

• Expertise in post-training of large language models, including:

• Policy optimization algorithms (e.g. GRPO, PPO)

• Verifier-based RL frameworks (RLVR)

• Supervised fine-tuning (SFT)

• Data-centric AI methods, including synthetic data and curation

• Parameter-efficient fine-tuning (PEFT)

• Preference and Safety Alignment

• Experience of driving research projects, delivering valuable findings and prototypes, and then converting them into products.

• Excellent communication skills in English and a talent for explaining complex scientific topics clearly and concisely.