Member of Technical Staff - Research Intern

What you'd do

• Responsible for co-designing and implementing the Reinforcement Learning experiments (GRPO/PPO/DPO), training data mixes and reward signal explorations.
• Contribute to research on post-training techniques, running ablation studies to improve model reasoning and alignment capabilities.
• Implement and test new algorithms for model fine-tuning and evaluation, helping to translate research papers into working prototypes.
• Analyze experimental results and debug model behavior to help establish best practices for our training recipes.

What they want

• Education: Currently pursuing a PhD or Master’s degree in Computer Science, Machine Learning, Mathematics, or a related field. Exceptional undergraduates with strong research experience are also encouraged to apply.
• RL Knowledge: Strong academic understanding or project experience with Reinforcement Learning (e.g., PPO, DPO, GRPO). You should be comfortable reading and implementing concepts from recent research papers.
• Coding Proficiency: Strong proficiency in Python and deep learning frameworks (PyTorch). You should be able to write clean, efficient research code.
• Research Mindset: A fast learner who is comfortable navigating ambiguity. You enjoy analyzing complex problems and iterating quickly on experiments.
• LLM Familiarity: Experience with training or fine-tuning Large Language Models (LLMs) or familiarity with the modern NLP stack (Transformers, HuggingFace, etc.).
• Previous internship experience at frontier AI labs or research organizations.
• Publications (or submissions) in top ML venues (NeurIPS, ICLR, ICML) or EDA venues (DAC, ICCAD).
• Familiarity with hardware design concepts (Verilog, RTL, EDA tools), though not required.

What you get

• Competitive internship stipend
• Mentorship from a team of researchers and engineers from Anthropic, DeepMind, Meta, and Stanford
• Opportunity to work on 0→1 problems in AI-driven chip design

About Architect Labs · Machine Learning

Architect is a frontier AI lab for custom silicon. We partner with frontier labs, clouds / neoclouds, physical AI companies, and advanced fabs to tape-out custom chips co-designed for next-generation AI workloads. Our goal is to compress end-to-end software to silicon timelines, and maximize intelligence per watt and per dollar for the world. We are a small exceptional team across silicon, systems, software and frontier AI.