Arboreal AI is building the next generation of tools to empower personalized wellness and positive habits.

Ginger and Headspace merged in late 2021 to become Headspace Health and I was invited to run AI and machine learning for the combined company.Headspace Health aims to help every human being on the planet with their mental health needs and deliver care at all levels of the spectrum: * Helping those who are thriving to maintain through targeted meditation and other self-care techniques * Meeting people who are coping where they are at with real time text-based behavioral health coaching* Lifting up those experiencing struggle, trauma and chronic conditions through video-based clinical practices and asynchronous followup.The Artificial Intelligence and Machine Learning (AI/ML) team, which I ran, built the next generation of ML, NLP, AI and analytics tools on top of an extraordinarily rich dataset in order to amplify quality of care and scale this care to the entire world.My team had deep collaboration with product, ops and other functions and I advised the executive team, educated leadership in multiple functions on AI as well as ran full-company breakouts on our companies state-of-the-art tech and trajectory.

Ginger's mission was to make high-quality mental and behavioral health support universally accessible within seconds of need. My team explored and built novel data products using machine learning, large language models (LLMs), and advanced analytics to deeply personalize member experience, unlock actionable understanding and upgrade quality of care to unprecedented levels, while accelerating scalability.I built the Data Science team out (from 2 to 21) with an inspired community of bright, ambitious machine learning engineers, analysts and data engineers eager to innovate in support of Ginger's mission: "a world where mental health is never an obstacle."

Number 4 on the founding team. Headed Core's algorithmic intelligence, cloud infrastructure and data insight evolution on the journey to creating a personalized ECG-based meditation trainer. Designed and implemented modern, cost efficient, and massively scalable cloud stack for high-performance API, data processing and multimedia delivery operations. Developed SQL data dashboard solution for 1/10th the price quoted elsewhere. Designed protocols for signal-chain quality analysis and machine learning data labeling. Vetted options for on-board ECG chip, quantifying accuracy while working with chip maker to improve RR-interval reporting precision by 16x. Improved product-wide issue visibility and response time through a firmware-to-cloud logging and alert solution with 100% coverage in the field. Developed A/B testing infrastructure and employed it to amplify customer engagement via new features. Core was acquired in mid-2021 by Hyperice as the start of its Mind product line.

Delivered a range of services for several clients:* Machine Learning expertise (15+ years) - Algorithms and end-end systems. Regression, classification, optimization, etc.* Data science - Turning imperfect large multi-dimensional datasets into insights. Feature extraction, analysis automation.* Big data engineering - Back-end processing, automation, algorithm and storage pipeline solutions via e.g. AWS, Spark, Streaming (Kafka, Kenisis, Spark Streaming, ...), Postgres/Redshift, HDFS.* Product Design - product feature choice, design and testing (A/B, experiment design)[note 2019: currently only available for light-touch advising roles for strongly aligned endevours]

Lead a team of 7 in iterating a progression of ideas for applications of ECG, accelerometer, and adaptive content delivery for wearable and phone-based apps. Aim was to deliver a user real-time access to their inferred pattern of cognitive-emotional state and to present relevant opportunities for positive change. Built upon our own prior work on a self-contained exhibit at the Exploratorium.

Project lead building production-level algorithms for distributed streaming accelerometer, heart rate (ECG) and event data analysis to serve health care providers and patients (FDA regulated). Transitioned data science team from a prototype-then-hand-off paradigm to coding directly in Python (streaming, Spark-ready), improving process efficiency and iteration time. Designed, implemented, optimized and brought to production a novel sleep categorization algorithm which was 1200% more accurate than previous. Improved and validated algorithm via IRB approved study, with results later presented at Sleep 2016. Collaborated with firmware engineers to upgrade input signals; worked with systems team to discover, characterize and address faults and dropouts in data. Mentored/managed several employees and interns. Lead Python instruction for the whole company.

Lead team of 4 in designing, building and iterating a data-rich exhibit booth. Exploration through simultaneous physiological and subjective signals of conscious influence of brain states via intentional practice. Using Python, Spacebrew (client/server), and JavaScript (D3) on top of InteraXon Muse EEG and NeuroSky ECG (heart rate variability) hardware. See http://explorecogtech.com and the Exploratorium press release here: http://www.exploratorium.edu/press-office/press-releases/new-exhibition-understanding-influencing-brain-activity-opens

Designing and building core infrastructure of the data science platform for analysis, adaptation and real-time optimized control of tens of thousands of homes, each with minute level data on several internal and environmental variables. Developed primarily in Python working with a three-man team managed in the style of Agile / Scrum. Designed methods for daily and monthly aggregation of finer resolution data. Orchestrated and managed map/reduce jobs on Hadoop/Hive/HBase. Further ad-hoc data analysis via MySQL queries, scripts and data visualisation. Note: position was part-time (30+ hours/wk) and concurrent with my doctoral work.

Research and work in data mining, adaptive control, pattern recognition and autonomy. Developed methods including time delay embedded regression using neural network ensembles, Gaussian processes and linear/quadratic methods for supervised learning. Developed a parameter space search framework for optimizing regression algorithms, parameters and data subset selection on a given set of data. Most work in Matlab, with some C, C++, Java, Perl throughout.

Research and work in machine learning, pattern recognition, biologically inspired computing, graphical and tactile user interfaces and experimental study design. Most work in Java, with some C++, Perl, Tcl/Tk throughout.