Machine Learning Engineer - Model Distillation and Knowledge Distillation

About the Role

This position, based in Noida or Bengaluru, involves developing efficient small-scale machine learning models using distillation techniques. The primary focus is on engineering and maintaining distillation pipelines that enable smaller models to perform on par with much larger counterparts while reducing computational costs.

Key Responsibilities

• Design and operate teacher-student distillation workflows to create compact, optimized models.
• Apply quantization-aware training and other model compression methods to enhance efficiency.
• Oversee iterative updating cycles of student models as teacher models and datasets evolve.
• Assess distilled models using benchmark tests and practical production scenarios to ensure quality.
• Work closely with evaluation and data teams to generate synthetic datasets supporting the distillation process.

Required Qualifications and Experience

• At least five years of experience in machine learning engineering or applied research fields.
• Practical expertise in knowledge distillation and teacher-student training paradigms.
• Skilled in implementing quantization-aware training and various model compression techniques.
• Proficient with PyTorch and Python programming.
• Ability to rapidly iterate on model trade-offs under computational constraints.

Educational Background

Master's or doctoral degree in Machine Learning, Computer Science, or a closely related quantitative discipline.

Preferred Additional Qualifications

• An advanced degree (Master's or PhD) in Machine Learning with published research in model compression or efficient inference.
• Experience in developing small but production-ready language models.