Dylan Jackfrie Email & Phone Number

* Led and mentored a team of four junior engineers in designing and implementing a data pipeline for AWS DeepComposer, a musical keyboard and service for exploring Generative AI. Made product ready for a Re:Invent 2020 preview by (1) leading a team in development of GAN-trained models, (2) leading a second team on operational readiness for the ML training and inference infrastructure, and (3) mentoring a peer on testing service reliability under load. Led development of educational content and ran Re:Invent 2020 workshops for DeepComposer.* Advised design on user experience during model training, defining domain-specific metrics demonstrating model convergence. Ran an org-wide technical book club to increase the group’s familiarity with Generative AI. Led team in releasing two new machine learning models for AWS DeepComposer (Autoregressive CNN-based architecture, and Transformer architecture), and in releasing a paper on GAN-style training of Transformers. Ran operations for maintenance of AWS DeepComposer (2021-2022).* Ran as technical lead of DeepLabs, an internal team dedicated to exploring new products under AWS AI Devices. Led engineering design and collaborated with a Senior Principal Applied Scientist in science de-risking for an unreleased product involving Transformer-based deep learning models and Biofeedback.Redefined the AI Devices new product ideation pipeline. Led team in organizing informational sessions bringing the org up to speed in popular emerging topics in AI/ML including GNNs, XAI, NLP, Causality, and Metalearning. Orchestrated brainstorming sessions across AWS AI Devices and affiliated internal orgs, and presented top ideas to VP of Artificial Intelligence at AWS.* Wrote coursework for the AWS Machine Learning Foundations Course on Udacity, including defining all lecture material for sections on AI/ML fundamentals and provided final review for all other content.

East Palo Alto, California, United States

* Led completion of frontend React code for GA of AWS DeepRacer, a 1/18th scale race car for exploring Reinforcement Learning. Advised product/design on technical language and information architecture used in the application. Advised cloud services team on schema design. Defined design for GDPR compliance.* Advised product on educational curriculum for an unreleased product for project-based introduction to machine learning. Designed device software packaging and deployment for this product.

San Francisco, California, United States

* Service development in Java and Ruby. Provided platform support for web service and async job backend (>200 job types), Redis infrastructure (10-13 clusters), MySQL infrastructure (~300 tables across 5 instances), Memcached infrastructure, and additional DynamoDB-backed services written in Java and Ruby (7 services).* Implemented a robotic traffic detection and mitigation system including the core ML model, data storage for datasets and models, and a web frontend for customer service to manage new threats.* Managed the Goodreads Choice Awards 2016 campaign as technical lead, reimplementing a portion of the feature in React.* Designed pattern for evaluating balancing of SQL tables across databases, and used this to plan the migration of ~300 tables from self-managed MySQL to AWS Aurora. Defined “virtual databases” and a system for managing ownership of data and code to better divide ownership of monolithic infrastructure across the company’s engineering teams.* Organized stability/scaling preparation for peak traffic in 2017, 2018. Created and ran a new process for regularly load testing, across 4 engineering teams and 12 services. Designed and implemented load testing infrastructure for simulating web traffic. Created system to continuously forecast peak traffic.* Built a data lake in AWS S3, PySpark, AWS Glue. Organized a company-wide “Machine Learning and Big Data” hackathon. Prototyped reimplementing book recommendations engine on this system.

San Fransisco

* Created reusable pattern for offloading MySQL record count queries to Redis, standardized code interfaces for Redis access, and improved Redis availability by introducing Redis Sentinel.* Decreased outage response time by adding health monitoring and alarming around core infrastructure.

Berkeley, Ca

Parkify was started by UC alumni who hated parking near campus and in the city. The staff at Parkify are experts on connecting spot owners with desparate drivers looking for valuable parking spaces. The parkify app is now available on the iTune app store and will be available on Google play shortly. Parkify is live at UC Berkeley and UCLA with several spots for rent near campus and downtown.

Improvement of computer security education, including making better labs and teaching material.Development of the Fall 2012 Intro to Computer Security Course.

San Luis Obispo, California Area