6 months ago
Stockholm, Sweden or London, United KingdomSenior
Responsibilities
- Own the full lifecycle of Lovable’s post-training pipeline from data curation and training runs through evaluation and production deployment.
- Apply reinforcement learning, preference optimization, and supervised fine-tuning to improve code generation, reasoning about user intent, and reliable agent behavior.
- Build evaluation and experimentation infrastructure covering helpfulness, safety, latency, reliability, and user satisfaction.
- Develop and operate production systems for large-scale training jobs, including GPU orchestration and data pipelines.
- Collaborate with agent, product, and infrastructure engineers to turn model improvements into product improvements.
- Investigate and resolve failures across training recipes, data, inference, serving, and production behavior.
- Read research papers, run experiments, and rapidly move promising research into production.
Requirements
- Personally run post-training jobs on large language models using methods such as RFT/RLVR or preference optimization.
- Write reliable production code and build systems that operate beyond research prototypes.
- Be fluent in at least one major machine learning framework, such as PyTorch or JAX.
- Have experience with distributed training setups and GPU clusters.
- Understand the mathematics of preference optimization, reward modeling, and alignment techniques.
- Have built or substantially contributed to evaluation systems measuring real-world model quality.
- Be able to trace model-quality regressions from user-facing symptoms through serving, inference, and training.
- Preferred qualifications include experience with code generation or agentic workloads, deploying post-trained models to real users at scale, owning data-to-deployment workflows, speculative decoding, predictive evaluation methodology, open-source ML contributions, or relevant publication experience.
Tech Stack
Categories
AI ResearchML Engineering
About Lovable
Lovable builds an AI-native software creation platform that lets people build apps and websites by chatting with AI, along with tools such as gptengineer.app. It serves solopreneurs, small businesses, and enterprise teams, and users have created over 60 million projects since its November 2024 launch, with customers including Adidas and Zendesk. The company is privately held and headquartered in Stockholm.