Machine Learning Engineer Intern
Commitment Period: Jan 2027 onwards (minimum 6 months, full time)
Job Description
- We are looking for a ML Engineer Intern to research and build methods that find best compressed neural networks. Focus of this work includes knowledge distillation (KD) and quantization-aware distillation (QAD).
- Deliverable: a comparison of techniques by model family, with reported recommendations.
Responsibilities
- Develop KD and QAD approaches across different model types
- Ablate key choices among various techniques
- Build reproducible training and evaluation scripts, and report findings
Techniques You Will Work On
- Knowledge distillation: soft-label, feature-based and self-distillation
- Quantization-aware training (QAT) and QAD
- Low-bit quantization with post-training quantization (PTQ) as the baseline
Requirements
- Currently studying CS, Engineering, AI/ML or a related field
- Commitment for at least 6 months full time
- Strong Python and hands-on PyTorch (custom training loops and losses)
- Trained and evaluated vision models end to end
- Knowledge of KD and quantization basics (temperature, PTQ vs QAT)
- Careful, reproducible experiments and clear written reports
Preferred (Plus)
- Project or thesis on model compression (quantization)
- Experience with PyTorch quantization tools or ONNX export
- Publications at leading conferences (i.e. CVPR, ICML, NeurIPS etc.)