Senior Machine Learning Scientist - Applied Research (USA Remote)

Machine Learning is integral to the continued success of our company. Our product roadmap is exciting and ambitious. You will join a global team of curious, helpful, and independent scientists and engineers, united by a commitment to deliver cutting-edge, well-engineered Machine Learning systems. You will work closely with product and engineering teams across Turnitin to integrate Machine Learning into a broad suite of learning, teaching and integrity products.

We are in a unique position to deliver Machine Learning used by hundreds of thousands of instructors teaching millions of students around the world. Your contributions will have global reach and scale. Billions of papers have been submitted to the Turnitin platform, and hundreds of millions of answers have been graded on the Gradescope and Examsoft platforms. Machine Learning powers our AI Writing detection system, gives automated feedback on student writing, investigates authorship of student writing, revolutionizes the creation and grading of assessments, and plays a critical role in many back-end processes.

Responsibilities and Requirements

We’re an applied science group leaning towards modern Deep Learning. We expect our Senior Machine Learning Scientists to have a well-balanced set of skills, both in the Science as well as Software Engineering aspects of (Deep) Machine Learning. You will focus on developing novel and deployable ML models and solutions where no ready-made solution may be available. Therefore you need to be conversant enough with the mathematics of machine learning and deep neural networks such that you can construct novel model architectures, loss functions, training methods, training loops etc. You are also expected to keep abreast of the latest research advancements in AI and Deep Learning across modalities and apply those to your work. While we leverage ready-made training platforms, we also write our own training loops. Additionally, the models need to be directly deployable in our products, therefore, production level coding and software engineering proficiency is required. You may train large models (up to 100s of billions of parameters) therefore, ability to train on multiple GPUs and nodes and knowledge of the latest model training and inferencing advancements is necessary. Next, the models must perform well in production not only in terms of accuracy but also compute-cost. Delivering such software requires a sufficiently deep Computer Science background. Dataset exploration, generation (synthetic), design, construction and analysis, are a routine part of the job and may occupy a significant fraction of your time. Also, datasets can be large (billions of samples), therefore the ability to write parallel and efficient pipelines is a necessary skill. You will also be involved in code & model maintenance, code hardening (preparing the model and code for production pipelines), developing and staging demos and presenting your work within the company as well as via publications in peer reviewed venues (preferably A/A+ rated).

Day-to-day, your responsibilities are to:

• Research and develop production grade Machine Learning models as described above. Optimize models for scaled production usage.
• Work with colleagues in the AI team, other Engineering teams, subject matter experts, Product Management, Marketing, Sales and Customer support to explore ongoing product issues, challenges and opportunities and then recommend innovative ML/AI based solutions.
• Help out with ad-hoc one-off tasks as a team player within the AI team. 
• Work with subject matter experts to curate and generate optimal datasets following responsible data collection and model maintenance practices. Explore and access SQL, no-SQL and web data and write efficient parallel pipelines. Review and design datasets to ensure data quality.
• Investigate weaknesses of models in production and work on pragmatic solutions.
• Utilize, adopt, and fine-tune off the shelf models, including LLMs exposed via API (through prompt engineering and agents) and locally hosting LMs and other foundation models.
• Stay current in the field - read research papers, experiment with new architectures and LLMs, and share your findings.
• Write clean, efficient, and modular code with automated tests and appropriate documentation.
• Stay up to date with technology and platforms, make good technological choices, and be able to explain them to the organization.
• Work with downstream teams to productionize your work and ensure that it makes into a product release.
• Communicate insights, as well as the behavior and limitations of models, to peers, subject matter experts, and product owners.
• Present and publish your work.