ML Researcher - Intern

Pareto Labs is building the learning layer for AI systems—infrastructure that enables models and agents to learn from real-world signals, organize memory, evaluate changes, and improve safely over time.
We are looking for a Machine Learning Research Intern to explore new methods for continual learning, post-training, model evaluation, and safe adaptation.
What you’ll do
Research methods for continual learning, post-training, and model adaptation
Design experiments for supervised fine-tuning, preference optimization, and reinforcement learning
Build evaluation frameworks for reliability, reasoning, safety, and domain performance
Investigate memory architectures and feedback loops for AI agents
Create and curate datasets from human feedback and real-world outcomes
Reproduce and extend relevant research papers
Analyze experimental results and communicate findings clearly
Collaborate with engineers to turn promising research into working prototypes
Who we’re looking for
Student or recent graduate in machine learning, computer science, mathematics, statistics, or a related field
Strong understanding of machine-learning and deep-learning fundamentals
Proficiency in Python and experience with PyTorch
Familiarity with transformers, large language models, and modern post-training methods
Experience designing experiments and interpreting quantitative results
Ability to read and implement ideas from research papers
Strong technical writing and communication skills
Familiarity with distributed training, reinforcement learning, or model evaluation is a plus
You may be a strong fit if you
Enjoy working on open-ended research problems
Care about both research quality and real-world deployment
Are curious about how AI systems can learn continuously without degrading safety or reliability
Move quickly, document your work, and challenge your own assumptions
What you’ll gain
You’ll work directly with the founding team on research at the intersection of continual learning, model post-training, memory, evaluation, and safe self-improvement. Strong work may contribute to technical publications, open-source research, and production systems.
Apply directly through LinkedIn using the Easy Apply button.

Work arrangement
No