About us

We’re on a mission to transform spoken communication for individuals, teams and organizations of any size. Meetings may be our most information-rich channel for work, but suffer from a lack of structure and documentation. At Supernormal, we’re solving this problem with focus, design and craft.

We’ve been working on this since 2019 and have customers like Snap, Salesforce, Replay, Gitcoin, Pinterest and thousands more on this journey with us. Today, we are growing rapidly and are excited for new teammates to join who are the best at what they do. We’re passionately building a team that is as diverse and creative as the millions of people we serve worldwide.

Supernormal is a remote-first company and does not require co-location. We have annual team retreats and gather quarterly.

About the role

Machine learning engineers at Supernormal build the AI that superpowers the core product experience for people’s meetings including transcription, note generation, and task automation. The AI team builds reliable and secure services that use the most advanced AI models in the market to generate millions of high-quality meeting notes to a rapidly growing customer base. Our work revolves heavily around software engineering, too – we are looking for people with a drive to roll up their sleeves and get new models and features out to users as quickly as possible.

What you’ll work on

As an ML engineer with a focus on modeling at Supernormal, you will specialize in developing and refining models that power our AI solutions for meeting notes, question answering, and task completion. Your expertise will be critical in LLM API calls, custom model training and deployment, speech recognition, quality evaluation and fixes, and retrieval augmented generation. You will primarily focus on optimizing model performance, cost, latency, and quality. Some of the projects you'll work on include:

• Prompt engineering using state-of-the-art techniques to improve the core meeting assistant scenarios.
• Building and shipping custom machine learning models to augment the AI stack, including improving transcript quality, reducing tokens sent to APIs, removing defects in LLM output, and extracting semi-structured data.
• Training and deploying custom large language models from open source using state-of-the-art techniques (LoRA, RLHF, instruction-tuning, etc).
• Specializing in model optimization to enhance performance and efficiency.
• Developing new product experiences using NLP & LLMs that get better based on user feedback & iteration while collaborating with product engineers & design team.
• Defining and improving business & product metrics to optimize the quality and cost of AI usage.
• Improving LLM-powered search and question answering (using RAG) over sets of meetings.
• Advocating for, and building, new and better ways of doing things. You’ll leave everything you touch just a bit better than you found it.