Research Engineering Manager - Model Training

📋 Description

• • Perplexity is at the forefront of revolutionizing how people access and interact with information, and at the heart of our innovative products lie our cutting-edge AI models. We are seeking a highly skilled and visionary Research Engineering Manager to lead our exceptional team of AI researchers and engineers. This is a pivotal role where you will guide the development of the core models that power Perplexity's premium, high-taste user experience, catering to some of the world's most discerning users.
• • Our team has already achieved significant breakthroughs in areas like agentic research and query understanding, developing models renowned for their accuracy and depth. As Perplexity continues its rapid growth, expanding our user base and product surfaces, the criticality of our in-house models escalates. You will play a direct role in shaping the future of AI-driven information discovery by diving deep into our vast and rich datasets of conversational and agentic queries.
• • Your primary focus will be on leveraging state-of-the-art training techniques to dramatically scale and enhance the performance of our AI models. This involves not just theoretical understanding but also hands-on technical leadership. You will empower your team to push the boundaries of what's possible, developing models that are not only technically superior but also directly address the most impactful use cases for our business and, most importantly, our users.
• • As the Research Engineering Manager, you will be responsible for leading a talented team of researchers and engineers dedicated to training state-of-the-art (SotA) models tailored for Perplexity-relevant applications. This leadership extends to driving research and engineering initiatives to bring production-ready models to life through advanced training and alignment methodologies, including sophisticated techniques like Reinforcement Learning (RL), Supervised Fine-Tuning (SFT), and other cutting-edge approaches.
• • A key aspect of this role is becoming intimately familiar with the team's technical stack. You will lead from the front, making significant hands-on technical contributions that demonstrate expertise and foster a culture of technical excellence. This deep engagement ensures you are well-positioned to guide the team through complex technical challenges.
• • You will own the entire lifecycle of our model development, including the critical data pipelines, training infrastructure, and evaluation frameworks required to train and continuously refine our Large Language Models (LLMs). This end-to-end ownership is crucial for maintaining and improving model quality and performance.
• • A significant part of your mandate will be to design and iterate on innovative model training and fine-tuning algorithms. This includes exploring and implementing preference-based methods, reinforcement learning from human or AI feedback (RLHF), and similar advanced techniques. Your approach will need to strike a delicate balance between scientific rigor, ensuring the robustness and theoretical soundness of our methods, and iteration velocity, enabling rapid experimentation and deployment in a fast-paced environment.
• • Furthermore, you will be instrumental in designing comprehensive evaluation strategies and continuously improving the production model training pipeline. The goal is to reliably deliver models that represent the Pareto frontier of speed and quality, ensuring our products are both performant and highly effective.
• • Close collaboration with our broader engineering teams is essential. You will work hand-in-hand with them to seamlessly integrate our in-house models into our product offerings. This collaboration will facilitate rapid iteration cycles, allowing us to quickly adapt and improve based on real-world usage data and user feedback.
• • Your responsibilities will also encompass the day-to-day execution, strategic project planning, and prioritization for the model training team. You will ensure the team consistently hits ambitious quality and performance goals, driving the team's success through effective management and clear direction.
• • This role offers a unique opportunity to shape the future of AI-powered search and information retrieval, working with a passionate team on challenging and impactful problems. If you are a leader with a deep technical background in AI and a passion for building world-class models, we encourage you to apply.