Christopher Kish Email & Phone Number

Pembroke, Ma, Us

* Created ML-team product infrastructure & processes with Databricks, Azure, and Azure AI Search * Built pipeline for fuzzy database record matching product using Spark, Databricks, a language model, a vector database, and a PowerBI monitoring dashboard* Created PoC system to find contextual word definitions using Spacy, LLama 3.1, and a vector database* Programmed several ETL pipelines in Databricks using both Spark and PythonReason for leaving: Part of lay-offs due to company having financial trouble

New Haven, Ct, Us

Fixed-term* Designed cognitive ability models for user behaviors on 2 games* Mentored & advised junior data scientistReason for leaving: Fixed-term

Berlin, Berlin, De

* Created company’s first machine-learning MLOps system to train and deploy a real-time model to populate a personalized list of songs using an in-memory vector database, Vertex, MLFlow, and Airflow* Wrote program to calculate feature values and send them to a feature store* Wrote program to retrieve features, train a model, and store the model* Analysed A/B tests of recommendation feature changesReason for leaving: Known short-term before joining. I wanted to move from Germany but had to wait for Covid restictions to be lifted.

Berlin, Berlin, De

* Team lead for selling data insights* Design machine learning tools for pricing, promotion evaluation, etc.* Designed web & mobile app scraping processReason for leaving: Re-org soon after joining resulted in having a job quite different to the one I applied for

Berlin, Be, De

* Invented & programmed a production pipeline of several machine-learning and Bayesian models to optimize advertising campaigns of 10,000s of individual jobs (the main product of the company, which raised the largest amount of money by a careers company in Germany)* Built a recommendation system with Spark and Tensorflow, combining deep learning, matrix factorisation, an autoencoder, and a BERT language model for >100% improvement* Created a system, incorporating doc2vec, to calculate optimal Facebook advertising audiences for 10,000s of individual ads, resulting in 40% reduction of cost* Created a classifier of various job attributes, benefits, and requirements from the job description using Hugging Face & BERT* Built a system to turn a job description into an embedding to be used by multiple models using SpaCy* Designed and built a complete data warehouse using AWS Redshift, Python and SQL ETL, Looker, Jenkins, Stitch, EC2, Git, and Snowplow tracking* Created multiple dashboards in Databricks using Spark & Looker to assess A/B experiments, monitor a model pipeline, and assess product and company performance* Wrote several automated, production programs to scrape data from websites and load it into Redshift* Designed & analyzed A/B experiments using Bayesian statisticsReason for leaving: Part of large-scale layoffs when company lost ~50% of revenue

Berlin, De

* Built a media mix valuation algorithm* Created a system to extract improvement points from customer survey free-text responses by building natural language processing tools* Programmed the computation & uploading of real-time bidding segments with Spark & machine learning* Built a machine-learning model to predict customer churn, improving performance 25%* Programmed many SQL data transformations to power reports, dashboards, and intermediate tables* Automated the loading and combining of data from Google Analytics, Appboy, DoubleClick, and email marketing to allow answering and monitoring key business questionsReason for leaving: Company dissolved my team (and several other technical teeams) due to a shift in business model

London, London, Gb

Developed machine learning solutions for a large organization to quicken knowledge acquisition and better predict outcomes of potential actions and participated in resulting sales meetings.Reason for leaving: Move to Germany

Hoboken, New Jersey, Us

* Led numerous projects designing analytics solutions to improve operations for both internal and external clients using a database of user actions with over 5B entries--quantified product efficacy, assessed performance of machine learning models of user behavior, and measured the impact of different content on user engagement* Developed a model for education publishing customers to estimate the quality of their assessment questions using only the student responses to those questions* Designed and implemented automated dashboards, one-off reports, and data services to answer key business questions, which involved collaborating with business development, product, marketing, and other teams; programming data ETL pipelines to populate a data warehouse; and analyzing data by creating MapReduce, Python, and SQL programs* Improved a statistical model for online prediction of user behavior that increased performance roughly 20%* Programmed data ETL pipelines to populate a data warehouse* Wrote MapReduce, Python, and SQL programsReason for leaving: I perceived significant risk (later borne out when the company went bankrupt) that we were significantly overselling our ability

London, London, Gb

* Advised team-leads through heads of a multi-billion dollar project by conducting ad hoc data analyses, building simulation models, developing metrics to manage the project, and forecasting future outcomes* Successfully predicted $2B of risk 2 years before it was realized for a $30B enterprise by building cost models from >1M data points* Contributed significantly to selling 3 $1-2M contracts* Developed strategy for complex, dynamic problems using a combination of data analysis and systems analysis by devising improvement actions, predicting risks multiple years in the future, and identifying key bottlenecks or inefficiencies* Created or collaborated on >100 presentations and deliverables for executive officers, managers, and technical teams* Identified the actions likely to produce the largest effect for an international counter-narcotics effort by building a Java program covering economic, political, military/police, and social forces accounting for 2nd- and higher-order effects* Devised management strategies for a software project that saved >$1M by constructing a data store, analyzing multiple data sources, and creating a model that accurately predicted software quality for >1 year* Designed a manufacturing purchasing strategy saving >$25M by analyzing shop-floor data and building a statistical model of the production process that accurately forecasted production rates for 2 years* Learned the underlying forces driving the gas extraction & use market in 2 weeks to create the sales slides for a $100B energy companyReason for leaving: Experience working internally

Cambridge, Ma, Us

* Informed policy decisions by building multiple models from millions of data points to predict the impact of global commercial aviation noise on housing and rent values* Built a statistical model to conduct Monte Carlo analyses on millions of data points totaling >$100B in real estate & rent value, uncovering billions of dollars of potential benefits from global airport noise reduction* Created regression & GIS models incorporating 8 population databases to predict housing values in ~80 regions of foreign countries across 5 continents* Assessed and utilized hedonic price models and willingness-to-pay survey methods to quantify the monetary value of environmental non-market goods* Conducted statistical analyses of satisfaction surveys to quantify the impact of highly-volatile variablesReason for leaving: Graduated

Manassas, Va, Us

Stanford, Ca, Us

M.S., Aeronautics & Astronautics

B.S., Mechanical Engineering

Education record