Primarily maintaining and extending Cyngn's vehicle simulations, including:

• maintaining and extending the simulation software

• working with QA to add new scenarios, and improve the daily simulation checks on master

• adding reporting capabilities that summarize simulation performance, in collaboration with autonomy engineers

I joined Nauto as #19, grew with the company to #190, so it was time for a break!

I direct the Driver Behavior Team, which creates Nauto's vehicle dynamics processing algos and risk algorithms, and provides data support for Nauto's insurance partners. Nauto is funded by Greylock Partners and Softbank.

• Expanded the Driver Behavior team from two to four extraordinarily talented algorithms experts.

• Took a nebulous scope of work, and defined a cohesive, progressive roadmap for the Driver Behavior team, addressing product needs with multiple algorithmic paths, each supporting the other.

• Led the team in executing this roadmap across five product release cycles.

• Coordinated with ENG teams and algorithms teams to define specs, standards for release, and process of deployment, as well as find opportunity for creative combinations of algorithms.

• Tracked and unblocked progress for my scrum team in daily stand ups and other meetings.

• As an individual contributor, developed and released a major revision to Nauto's risk scoring algorithm, in coordination with cloud and front end teams on new interactive UI elements displaying risk scores.

• Together with a fantastic engineer, created a major piece of infrastructure to enable algorithms to compensate for individual vehicle variation.

Public talks:

• Panelist, The Future of Telematics Risk Prevention, TU Connected Car, September 5th, 2017

I work in a dual role, first and foremost as a data scientist leading Nauto's risk scoring efforts. And second, in sales, building partnerships and acting as a technical liaison with leading insurers, insurance service providers, and insurance consultants. As a data scientist, I:

• Scoped, conceptualized, and designed four versions of Nauto’s risk estimation algorithm, VERA.

• Coordinated the efforts of seven staff across Android, backend, frontend, data science, and design teams to implement VERA’s data collection, implementation, and UI/UX.

• Designed, trained off bootstrapped data, and implemented engineering to deploy VERA scoring models in production. Principal tools: Python, sci-kit learn (sklearn), Docker.

As a sales and implementation manager, I:

• Participated in or lead over 100 partner pitches and discussions, including top tier insurers.

• Collected feedback and requirements from potential partners on VERA.

• Design and negotiate proposals for partnerships, including pilot scoping and pilot evaluation criteria.

• Very occasionally meet informally with investors and board members to discuss risk assessment.

One of only 46 fellows chosen from 1,807 applicants to complete rigorous seven week data science curriculum.

• Completed 34 homework assignments covering Python, web scraping, data munging, XML parsing, Pandas, Bokeh, Flask, and other libraries.

• Programmed jobs on AWS using both MapReduce (MRJob) and Spark with Scala.

• Used the scikit-learn Python library to do high dimensional linear regression, K-means, and other forms of machine learning.

• Worked with NTLK to analyze n-grams in combination with MapReduce.

• Munged and analyzed 6 million+ collision records for California for my capstone project: http://onebilliondollars.herokuapp.com/