Jonathan Conley Email & Phone Number

Mountain View, Ca, Us

Artificial Intelligence and Machine Learning within Google's Syndicated Text Ads team.Highlights:Primary lead on several machine learning research projects for the syndicated text ads team.1. Solo researcher on a project that reduced the amount of training data necessary and training time by 75% for our neural networks. This involved tuning, investigations, and piecing together checkpoints from models that trained on separate data. The end result allowed enough training headroom to continue to grow our team's models.2. Trained variational auto encoder's using categorical vectors as the target output of a machine learning model to help the team cluster advertiser domains based on the type's of ads that lead to that domain.3. Trained our teams first neural networks (2016) and tuned for accuracy that drove a $XXM increase in revenue.One of two primary drivers to launch text ads on Youtube Search. The initial launch was worth $XXX million a year. This included additional machine learning modeling to account for a new large traffic source.Lead on a practical LLM project for the use case of related searches based on the content to LLM responses. This involves initial neural network architecture, serving code, and logging data.

Designer, Server Architect, Java 7 Engineer, Web developer, App Developer, Database admin, and Promoter of MyFlixFinder.com.

Orlando, Florida, Us

Physician Skill Decay Prediction – Worked with and under data scientists to create an expert system using a combination of Bayesian statistics, clustering, and the rules engine Jess to determine when a physician’s skills have decayed. Designed a clustering method and implemented the ruleset, Bayesian component, and a Java server that was used for communication between elements. Also wrote and optimized queries to improve runtime of the different components by weighing database access time vs. extraneous data causing slowdowns.Amputee Training Games – Designed and implemented a suite of training games controlled using a Bluetooth myoelectric armband. All games were designed to use muscle movements that mimicked ones used to control a myoelectric prosthesis. The project involved writing native Android libraries to gather the electromyography (EMG) data from the armband and creating threads that used queues to help process the data and send it to the games.Virtual Reality Study – Wrote software to collect reaction time and 3D movement data, and a server to handle the data for a study that determined the physical effects of a VR headset after multiple hours of use. Ran participants through the study by connecting biological sensors and then running them through variations of reaction time, steadiness, and visual acuity tests.Stress Classification and Rating Application – Developed an Android application that connected to a wristband via Bluetooth to collect Blood Volume Pressure (BVP) and Electrodermal Activity (EDA). Implemented an algorithm to turn BVP into heart rate, then used the heart rate and EDA as input into a classifier to determine how stressed an individual was. Used HTTP requests and binary data to send raw data to a server for further analysis. Tuned said server to allow quick inserts and access to a table that contained hundreds of millions of records.

Orlando, Us

Front facing server- Wrote Java code that was used to communicate through HTTP requests with a mobile client or website. Also wrote modules used to communicate with other Java services in response to HTTP requests.Interest Matcher – Worked in tandem with a specialist to develop an algorithm to match events to users based on the interests mined from their Facebook data.Artist Resolving Service – Created a service to retrieve music file names and then use common characteristics in the form of a regular expression to split the filenames and then do a lookup into a table of known artists to determine what artist performed the music.