Principal Machine Learning Scientist, Turnitin

• 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
• 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
• 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
• Master's degree or PhD in Computer Science, Electrical Engineering, AI, Machine Learning, applied math or related field or outstanding previous achievements demonstrating excellence in Deep Machine Learning, Computer Science and Software Engineering
• At least 10 years of industry experience in Machine / Deep Learning (we use the python ecosystem for ML), Computer Science and Software Engineering
• A strong understanding of the math and theory behind machine learning and deep learning is a prerequisite
• Academic publications in peer reviewed conferences or journals related to Machine Learning - preferably A/A+ rated such as NeurIPS, ICML, ICLR, AAAI, TMLR, JMLR, IJCAI, ICANN, KDD, ACL, EMNLP, NAACL, COLING, CVPR, ICCV, ECCV, IEEE etc
• Machine / Deep Learning development skills, including popular platforms (we use AWS SageMaker, Hugging Face, Transformers, PyTorch, PyTorch Lightning, Ray, scikit-learn, Jupyter, Weights & Biases etc.)
• An understanding of Language Models, using and training / fine-tuning and a familiarity with industry-standard LM families
• Excellent communication and teamwork skills
• Fluent in written and spoken English
• At Turnitin, we recognize it’s unrealistic for candidates to fulfill 100% of the criteria in a job ad. We encourage you to apply if you meet the majority of the requirements because we know that skills evolve over time

Desirable

• 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
• You will focus on developing novel and deployable ML models and solutions where no ready-made solution may be available
• 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
• 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 developing and staging demos and presenting your work within the company as well as via publications in peer reviewed venues (preferably A/A+ rated)
• You'll research and develop 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 local datastores as well as 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
• Modify and fine-tune off the shelf models or develop novel models. Use LLMs via API (through prompt engineering and agents) and locally hosted LLMs and other foundation models
• Stay current in the field - read research papers, experiment with new architectures and methods, 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