James Kelleher Email and Phone Number

Regular additions and enhancements to existing core ETL pipeline, which consists of several persistent microservices written in Python and Scala, connected via Kafka, hosted in AWS via CDK with data stores in Dynamo and MongoI led a project to deliver aggregates on large datasets in realtime to customers. I explored several solutions, ultimately choosing Pinot for its flexibility and fast response time, as well as its support of streaming upserts for extra quick post-load turnaround times. I set up a working prototype using Kubernetes and CDK(Note that I was hired at a company called Helm, which Murmuration acquired in 2023)

Created jobs in Airflow to deliver data to our internal Snowflake database and to external APIs.Made modifications to our Airflow's Kubernetes cluster to improve developer onboarding and productivity.Onboarded Postgres databases to AWS DMS.Updated and fixed Python jobs that handled data migrations to Snowflake.

I primarily worked on ETL pipelines for several of our internal applications, using a mixture of Python and shell scripting. These had been manual processes that took hours to complete, so I made architectural changes to convert them into processes that could be installed and ran easily on a remote machine. I added Slack integrations to report on pipeline status and to raise any possible errors in a public fashion. Additionally, I am currently working on getting these processes running in AWS.I prototyped a new internal application, where I created a new daily ETL from GraphQL to Python to an AWS Postgres database I setup, and displayed the data in Metabase. The ETL was kept for the final application. I am maintaining the ETL and updating it for any changes happen in the source data.I performed several research and analytics tasks, such as the identification of records that had been duplicated in our database, and quantifying market coverage for our CPO.

At CraftJack, I used Python, Pandas, and SQL (Snowflake) to analyze auctions of contractor leads between lead generation networks, and performed data modeling to explore new bidding strategies to improve profitability. I experimented with several different models and selected the model that performed best with the scale and nature of our data. I had regular communication with the CEO, and wrote up many of my findings as Jupyter notebook reports, taking care to build a business case and communicate the argument in a way understandable to someone without a math/CS background.I also spent time investigating CraftJack's SEM strategy, identifying keywords that were underperforming and analyzing potential causes for the downturn.

I was in charge of the design and implementation of Prescriptive Data's analytics microservice infrastructure. Using Python and Pandas, I built a suite of programs to provide useful, actionable metrics. As the team lead, I was in charge of hiring new data scientists and training them to perform in a startup environment. I managed our Agile workflow using JIRA and provided direction on challenging, ability-testing problems.

I architected and implemented an internal Python API to help our data science team quickly get clean data and perform useful analyses

Alongside Alex Leeds, I consulted Skyhorse Publishing on the technical requirements needed to develop their business intelligence

As a member of Conductor's QA team assigned to assisting the backend team I improved our data collection, ETL, and reporting pipeline using Java, S3, Hive, Oozie, and Spring. My tasks for the QA team itself included maintaining unit tests written in Java and frontend tests written in Cucumber/Ruby.

I helped PivotalPath better understand and utilize information they obtained from surveys. I established correlations and independence between survey questions, normalized for surveyor bias, and created measurements that are more objective, precise, and descriptive. I also created visualizations and reports to assist analysts in their research and decision making. I worked with Python and Pandas, a Python library that provides R-like functionality.

While on the Data Team at Huffington Post I worked on two major projects.The first was to write a Python script to gather tweets using Twitter’s streaming API and convert them to the same JSON format used by Firehose, an API that streams tweets at a higher rate, but for a monthly subscription with a data cap (the streaming API is free and has no such limit). Using my script, the data team could test their code without using up their monthly allotment and easily switch the input source over to Firehose when ready, ensuring that development would be possible at all times.I was also tasked with building a data visualizer to help the team better understand the behavior of weekly user returns. I decided that the visualizer should be interactive, portable, and scalable, and so taught myself D3.js and jQuery in order to implement the visualizer as a web application that runs in JavaScript. I also worked on improving one of the team’s machine learning algorithm by identifying patterns of error and suggesting possible solutions.

When Huffington Post updated the UI of the internal production side of their website, the majority of their unit tests – which navigate web pages using HTML and CSS – broke. As a QA intern, I rewrote dozens of these tests (built in Java using Selenium WebDriver) in order to help ensure site functionality. I also handled several trouble tickets a day, where it was my job to reproduce a reported error and provide the development team with any necessary information (which could be text, images or video) should an error be encountered.

I designed and built a game meant to help teach the Arabic language with a fellow student as part of an independent study.The other student was an international politics major with a strong passion for Arabic. We realized how algorithmic the language was, and how well suited it would be to an interactive, instructional format such as video games. Even though I knew no Arabic and she knew no programming, we determined that we could build a prototype anyway. I designed the overarching game and developed it using JavaScript and Processing, while she designed the lesson plans and implemented the Arabic features. We tested our prototype on fellow students and updated the game with their feedback.I am currently working on rebuilding the game in Unity + C# in my free time.

At the end of my summer at iD Tech Camps, I spoke with the parents of a camper who was dying to learn how to mod Minecraft, but was too young for the programming classes offered. I suggested that I teach him Java, the language Minecraft is written in, and they agreed. For five months I taught the student fundamentals of programming, up to the basics of OOP.

I conducted a weeklong class (for eight consecutive weeks) on intermediate Java, basic OOP, and modding the popular video game Minecraft (which is written in Java) to middle and high school students. During class time I would troubleshoot code, give detailed lessons on the intricacies of Java/OOP, and design projects tailored to each student's particular interests and skill levels so that they would leave camp each day with feelings of pride and accomplishment.