Will Lamond Email and Phone Number

Acquired by WHOOP in Nov '23Generative AI science and engineering, recommender systems, language modeling, and more.

2019-2020: Research + development of sleep stage and activity modeling for the Amazon Halo and Halo View band. Develop deep learning models for high precision sleep stage classification, and high performance models for embedded sleep detection on Halo devices.2021: I moved to Softlines, where I worked to automate the detection of apparel catalog data defects using machine learning. Engineered elastically scalable training framework using Pytorch + Ray. Implemented large-scale, self-supervised pre-training for image embedding models for use in downstream tasks.2022: Research + development of machine learning models for outfit recommendations. Leveraged curated outfit data to develop content-based recommender systems. Incorporated explicit and implicit customer feedback signals to improve recommendation quality.In addition to my project work, I was responsible for filing patents (when relevant) and giving talks at internal conferences on my team's work. I also reviewed papers for AMLC and ACVC.

Research supervised and unsupervised deep learning for MRI reconstruction in the Low Field Imaging Lab: http://martinos.org/lab/lfi Research applications of deep learning for recognizing genotypes of cancer driver mutations in whole slide pathology images.

Acquired by Meta.I led BlinkAI's R&D of machine learning software for digital cameras in low light environments. Responsible for: * Machine learning research (deep learning, computer vision, signal processing)* Product development (requirements setting + gathering, software engineering, product management)* VC (demonstrating capabilities + use cases during pitches)

Lead scientist and architect for Wayfair's computer vision group. Conceived and designed Wayfair's visual search platform, bringing it from a research project to a production platform serving external and internal users.* Design and train deep learning models for Wayfair's visual search platform.* Apply deep learning to a variety of computer vision problems. I utilize convolutional neural nets, transfer learning, and multi-task learning.* Develop high performance training systems for deep neural networks using state-of-the-art tools.* Advise other scientists and engineers on machine learning problems in computer vision, search, recommendations, personalization, and more.* Applied latent semantic indexing to recommendation tasks.* Developed a Bayesian network for learning customer preferences from observations in our clickstream. I based its design on algorithms from the information retrieval literature.* Built and deployed a Python web service for delivering personalized search results in a latency-sensitive application. It utilizes our Bayesian network user model and the LambdaMART algorithm for learning to rank.I also assist other engineers in a consultative manner, giving advice on architecture choices, high performance data structure selection, and general questions about the wayfair.com software ecosystem.

I work on the search and recommendations team. I regularly work in a Linux environment with Python, SQL, Hive/Hadoop, and occasionally bash script and PHP.I have made contributions to a wide range of systems, including the customer-facing website, text classification system, recommendation services, and personalized search. Examples include improving how a text classification system identifies features when the given text is German, identifying problems with and adjusting how the text classification results identify product types from keyword searches, designing and building the search engine for our flash sale site Joss & Main, and incorporating price-point data into our search engine to give personalized search results.

I worked on a team of peer reviewers for scientific papers submitted to SIGIR. I was responsible for evaluating submitted papers in terms of relevance to the conference subject, originality of work, technical soundness, and quality of presentation.

Assisted with teaching students during an 11 week session of General Assembly's Data Science course. * I helped students during in class exercises involving basic machine learning techniques, exploratory data analysis, data visualization, and general programming in Python. * I held office hours outside of class where I provided extra assistance. * I designed and delivered a tutorial on effectively using git.

I was responsible for developing, maintaining, and documenting Seniorlink's proprietary healthcare case management software. I started off developing new parts of the application in PHP and JavaScript, designing relevant portions of the database schema and developing the necessary interfaces between them. My responsibilities involved making changes throughout the entire application stack.I worked on a small team of software developers that are involved in the entire development process: we carefully discuss requirements, set ambitious yet realistic goals, rapidly prototype and test, and iterate on a tight schedule. I built a system, written in Python, for document organization that sorts input documents that follow a simple naming convention. The system places documents in the case management system for further use throughout the application, and saves staff members many hours a week from dealing directly with the necessary paper forms.I developed a Web crawler that fetches and organizes medicine data from the National Institute of Health. It works by querying the NIH's MedlinePlus RESTful Web service for relevant pages, parsing each page using the BeautifulSoup library, and then writing that data to our database for future use. I worked to develop a RESTful Web service to back Seniorlink's Android application. The service allows the exchange of case data in JSON, is written in Java using JAX-RS, and is deployed on JBoss 7.1.1 AS. It handles data through both a pre-existing SOAP service through a wsconsume-generated JAX-WS interface, and a DataSource connection pooling mechanism.

I worked on improving a framework for developing RESTful Web services and applications using Java servlets deployed with Apache Tomcat. The system uses feature testing and HTTP header inspection to deduce client capabilities with reasonable levels of certainty. It then uses its findings to present a representation of the requested resource in an appropriate way. Since the system was designed to not maintain client state, some portions of content that are static were computed on an as-needed basis. This works very well for presenting appropriate representations for each client, but in some cases also results in recomputing content that never changes. One of my major contributions is a specialized data structure that clusters and stores these snippets of content in such a way that they only get computed once, while still maintaining the RESTful nature of the framework.

I helped improved the performance of the University of Maine Ice Sheet Model (UMISM) by reorganizing the model architecture to allow a viewing application to receive model renderings directly from a rendering server instead of a connection server. UMISM is written in C and uses the Open MPI library.