James Thornton Email & Phone Number

Mountain View, Ca, Us

• Engineering project lead for high-security on device blocklists in Android Messages. This spam protection is now installed on more than 1 billion devices and is currently protecting over 800M active users.• Led system-health initiatives to evaluate on-device ML models during secure downloads and edge inference. • Led the low-tier and Go rollouts to ship on-device privacy preserving protections for over 150 million devices that were previously unsupported due to challenging hardware constraints.• Designed, owned, and built the first automated performance profiling infrastructure applied to spam detection in Android Messages.• Member of the Privacy Working Group for counter-abuse technologies; responsible for reviewing code changes and cross-referencing with legal documents to ensure the user's privacy is always protected.

Palo Alto, California, Us

• Led the machine learning research for Discover, a system that can automatically find new defects anywhere on the customer's product.• Led the Automatic Barcode Scanning Initiative: owned ML system design, led three engineers and one operations associate, researched computer vision algorithms, and deployed to factories in China on behalf of customers in Silicon Valley.• Led various initiatives involving ML research, ML infrastructure, distributed systems, and backend development.• Worked with Python, Scala, C, PyTorch, TF, deep learning, meta learning, SVMs, CNNs, anomaly detection, data visualization, Jupyter, numpy, scikit-learn, SQL, Docker, AWS (Kinesis, Lambda), Terraform, Bash, Linux, and TypeScript.

Emeryville, Ca, Us

• Trained neural networks to observe cells in biology experiments; increased the 20x zoomed in F1 score from 42% to 96%; increased the B Cell segmentation F1 score from 86% to 92%; trained a new Jurkat cell segmentation model with 99% F1 score; scheduled for production release.• Developed ML infrastructure increasing the amount of data that could be used in training by more than 6 times; built infrastructure for weighted loss functions; worked with Tensorflow, OpenCV, Numpy, and C#.• Supervised and trained three interns to form a team for human-in-the-loop data labeling.• Master's level internship; completed the project three weeks ahead of schedule• Spent the last three weeks prototyping the cell detection state of the art with Mask R-CNN.

San Francisco, California, Us

• DHS Passenger Screening Competition: Silver Medal - Solo Competitor - Placed 19th out of 518 teams in a national deep learning competition with a $1.5 million prize pool; built an extensive data pipeline from scratch; used a CNN ensemble and 27 neural networks in total.• NOAA Steller Sea Lions Competition: Bronze Medal - Solo Competitor - Placed 59th out of 385 teams in an international data science competition; utilized transfer learning and ensemble techniques; built a custom convolutional neural network architecture.• Built a deep learning server from scratch with a 1080Ti GPU and remote SSH access.• Reached the top 1% (highest rank: 685 out of 70,167) in a competitive community of machine learning experts, PhD holders, and professional data scientists.

Boston, Ma, Us

• Integral member of the team behind the big data precision medicine system that led to Genospace's acquisition by the HCA.• Developed multi-threaded backend infrastructure for the data pipeline and UI components for data preparation.• Worked full stack with technologies and concepts such as Java, Groovy/Grails, Backbone.js, Objective C, big data, object oriented design, and MongoDB.• Took ownership of the iOS User Analytics system; one of two engineers behind the entire iOS application.

Middlebury, Vermont, Us

• Tutored and graded for CS311: Artificial Intelligence, covering topics such as probability theory, Bayesian inference and text classification, Bayesian networks, breadth-first search, depth-first search, A*, constraint-satisfaction problems, and Markov models. (Java)• Tutored up to 30 students at a time from CS201: Data Structures on sorting algorithms, debugging strategies, and data structures such as stack, heap, trees, etc. (Java)• Tutored students from CS101 and CS150 on basic topics such as loops and recursion. (Python)

Boston, Ma, Us

• Owned and completed an extensive data visualization project, which is now in production and compares genetic and clinical patient attributes by displaying boxplots, linear regression lines, and tables based off of Chi-squared testing.• Implemented bug fixes, features, integration tests, unit tests, and database scripts.• Experienced challenges and shared responsibilities as part of a rapidly growing computational biology start-up.

• Led a team including two other programmers and a designer to develop an app that visualizes the user’s journal.• Built the word bubble visualization using D3’s Force Layout, optimizing collision detection with quadtrees.• Responsible for all Objective-C and Swift development.

• Worked with Objective C, Cocos 2D, the Javascript library D3, and Parse to develop iPhone and iPad applications.• Released four iOS applications: Memosee, Two Player Chess, FreezeDrop, and Sexile.• Over 15,000 downloads.

• Architected and implemented the email login system, handling concerns such as security, backwards compatibility, and client-server interaction.• Added social media support such as receiving Twitter feeds from the teams playing in each game, the ability to invite friends via Facebook, and Deep Linking for those requests to lead to a specific screen in the app.•Responsible for various graphics components, such as Implementing the CrowdSporting loading animation as a reusable class.

• Designed posters, t-shirts, and stickers for all MCAB events, such as a concert featuring Grammy Award winning “Fun.,” Middlebury’s Winter Carnival, and the weeklong Midd Mayhem festival.

Oxford, Gb

• Took ownership, within the team, of everything related to machine learning.• Researched applications of machine learning to demand forecasting in the supply chain; recommended the use of support vector regression and Gaussian processes based on the current state of technology; indicated that neural networks showed promise in the long term.• Worked with a five-person team of post- and undergraduates from the University of Oxford to consult for Tesco PLC, by profits the world’s second largest retailer; presented final report at Tesco world headquarters in Cheshunt, UK.

Middlebury, Vermont, Us

• Chosen as 1 of 12 students to receive college funding to consult for local small businesses.• Developed a general financial consulting report for the Aurora School, a local private elementary school; presented this to Middlebury College President Ronald Leibowitz.• Developed a customer feedback system for Noonie Deli using graphic design, HTML, and social media; used feedback to analyze options and make strategic recommendations such as offering free customer Wi-Fi.• As part of the MiddCOREplus program, participated in Friday workshops on accounting, businessmodeling, social media, graphic design, interviewing, and storytelling.

Us

• Conducted research through SEC’s EDGAR system to update Excel models analyzing the financial histories of companies in KDP’s portfolio and tracking the performance of KDP employed CFAs.

B.A, Magna Cum Laude, Rhetoric (Independent Scholar), Computer Science

Business Bridge Program

Middlebury Center For Medieval And Renaissance Studies, Keble College, Study Abroad