John Trenkle
Distinguished Principal Machine Learning Engineer @Tubi
Signup · Get unlimited contacts
WORK HISTORY
Distinguished Principal Machine Learning Engineer @Tubi
San Franciso, CA, US
As a seasoned expert in Recommendation Systems, NLP, and AdTech, I\'ve made significantcontributions to bridging the gap between content metadata and user understanding.My key achievements include:* Developing the \"Rich Content\" platform, which aggregates extensive metadata on movies, TVseries, and celebrities from both public and proprietary sources.* Creating the \"AutoData\" pipeline, leveraging LLMs prompts to extract robust insights thatenhance personalization and content comprehension.* Designing the \"Rosetta Stone\" system, which uses LLM-based embeddings and modern matchingtechniques to reconcile disparate ID spaces used in media.* Building the \"Embedding Menagerie\", a comprehensive collection of embeddings created fromdiverse sources, including word2vec/doc2vec, BERT, Transformer models, and custom DeepLearning frameworks.* Constructing the \"Container Genesis\" system, which automates the discovery and rendering ofrows on the Tubi Home Grid.* Developing the \"Rabbit AI\" semantic search capability using OpenAI GPT, capitalizing on theChatGPT momentum.* Addressing the \"Fan-of Modeling\" challenge to assess users\' affinity for various contenttypes.* Applying my expertise in cold starting, onboarding, content valuation, and recommendationsystems to improve user experiences.In my current role, I continue to drive innovation by generating new ideas and pursuing thosewith merit, aiming to stay at the forefront of the field.
EDUCATION
Wayne State University
BSEE, Electrical Engineering
SKILLS
ABOUT JOHN TRENKLE
I am a Well-Seasoned Pro in Machine Learning and Artificial Intelligence. I have worked in the field of NLP and have participated in its evolution from Symbolic Processing and Semantic Networks, to Hidden Markov Models and N-Grams, early Connectionist models, and more recently embeddings starting with Word2Vec (and many X-toVecs), BERT, and Transformers. These days, there is a new LLM each week. I\'ve seen it all, but know that there\'s more to come.I operate as a Principal Engineer (IC) and Tech Leader to solve problems in my domain du jour. Throughout my career, I have had the good fortune to work on diverse and challenging problems in Recommendation, Sensor Systems, Multilingual Text Recognition, Computer Vision, Drug Discovery, AdTech and more.I leverage my years of experience to rapidly develop testable prototypes, but my ultimate goal is to develop elegant end-to-end solutions to rewarding problems through initial concept, design, planning, implementation & deployment.I have handled Big Data, Small Data, Streaming Data and Sensor Data. I have worked with distributed computing from low-level multi-threading to Hadoop and Spark. I code predominantly in Python in DataBricks these days. I leverage GPUs when I can.I stay up-to-date on AI & Deep Learning literature and continually expand my skillset. To unlock epiphanies, one needs to broaden their knowledge. Every technique has its value – the right tool for the job. As machine learning practitioners know, there\'s no shortcut; every problem requires a viable solution.I keep my finger on the pulse of the Machine Learning world to know when new innovative approaches and packages come out. Whether TensorFlow or PyTorch, Transformers or Self-Supervised Learning, I read the literature, assess the potential, kick the tires and leverage the concepts and code that help solve current problems in my purview.What are my strongest qualities?• extensive and diverse experience• technical thought leadership• communicating complicated ideas simply• tenacity• modesty• the ability to think on my feet• saying “I don’t know” when I don’t• did I mention modesty?Despite the last one, here’s where I say good things about myself: I am a skilled technologist, sometimes a pioneer, sometimes applying other folks\' smart ideas in different arenas to new challenges. I believe that I can be a valuable asset to any company.List of publications:
This profile is compiled from publicly available professional sources. Unifers is not affiliated with or endorsed by LinkedIn. Request removal of this profile.