Machine Learning Science Intern

About the Role

Join distil labs as a Machine Learning Science Intern to contribute to revolutionary AI model efficiency. We are developing a platform that can transform a handful of seed examples or real-world data traces into compact language models that rival or surpass large cloud-based LLMs in accuracy, while operating privately on local or edge devices. As part of a small, impactful research team, you will collaborate with seasoned ML scientists and engineers over 4-6 months to innovate in knowledge distillation, synthetic data creation, and model self-enhancement, delivering prototypes, production features, and academic papers.

Key Projects

• Improve synthetic data diversity ensuring generated datasets accurately represent desired distributions without drifting.
• Enhance validation methods to determine the quality thresholds for synthetic training examples beyond current similarity and schema checks.
• Advance knowledge distillation techniques to create even smaller models (100M-1B parameters) without losing task accuracy, including multi-task and multi-modal approaches.
• Explore reasoning and tool usage capabilities to further empower smaller models on complex agentic tasks.

Responsibilities

• Conduct research and build prototypes for novel knowledge distillation algorithms that maintain accuracy while reducing model size.
• Expand internal benchmark suites to evaluate latency, memory usage, and task outcomes across diverse hardware.
• Collaborate with engineers to validate prototypes on real workloads and integrate effective solutions into Python/PyTorch systems and public APIs.
• Publish findings and present benchmark results at leading machine learning conferences.
• Engage actively in team culture through weekly demos, academic discussions, and product feedback.

Candidate Profile

• Enrolled PhD student in Computer Science, Machine Learning, Robotics, Statistics, Mathematics, or related disciplines.
• Strong foundational knowledge in machine learning along with proficiency in areas like deep learning, NLP, reinforcement learning, or optimization.
• Expertise in Python and at least one modern ML framework such as PyTorch, JAX, or TensorFlow.
• Capable of designing clear experimental plans to address open research questions.

Desirable Qualifications

• Authorship or co-authorship of research papers in workshops, conferences, or journals.
• Experience with model compression techniques (distillation, quantization, pruning), synthetic data generation, or self-improving models.
• Familiarity with large language models and LLM-centric research.
• Contributions to open-source ML projects and an active GitHub profile.

Why Join Us?

• Engage in cutting-edge research problems that impact industries like defence, robotics, and industrial sectors.
• Receive mentorship and direct access to founding scientists ensuring your professional growth.
• Supported opportunities to publish and attend conferences, including sponsorship for travel and registration.
• Competitive compensation exceeding market rates and coverage for travel expenses to Berlin offsite meetings.
• Enjoy remote-first work within European time zones, with quarterly in-person gatherings in Berlin.

Internship Details & Additional Info

• Flexible start date with rolling application reviews.
• Commitment of 4-6 months full-time.
• Remote work from anywhere in EU time zones, with occasional onsite weeks in Berlin quarterly.

Equal Opportunity Statement

distil labs is committed to diversity and inclusion, ensuring equal opportunity employment without discrimination based on race, religion, color, national origin, gender, sexual orientation, age, marital status, veteran status, or disability.