Phil Filippak

- Building internal tools for the team.

- Designing PC and board games as a hobby in my free time. - Learning how to work with game engines and libraries.

- Designed and built 3 graph note-taking apps, focusing on scalable data structure design. - Standardized the transactions model for collaborative editing in one of the applications. - Developed and integrated a semantic search layer for the flagship note-taking app. - Fixed and improved multiple UX aspects of the apps, including quick voice capture, infinite scroll usability, gesture controls for mobile devices, and many others.

- Led an ML team of 4 people and acted as a bridge between the analog chip engineering team and the ML one, arbitraging the exchange of requirements between the two. - Provided mentorship and guidance on a broad range of technical topics. - Achieved ~10x reduction in model size with only ~6% drop in MAE on a regression task and ~14% in F₂ score on binary classification which led to the 40% reduction of the final chip size. - Developed and maintained a service for demo previews of Keras models using Python and FastAPI with secure isolated instances focused on not leaking data for 5 different customers. - Built more that 40 reports for model performance on several datasets with Pandas & Pyplot.

Started working as a machine learning engineer and was swiftly promoted to Technical Lead.

- Set up the infrastructure using Django and PostgreSQL and a PyTorch-based ML pipeline. - Implemented a Python server app with 99.98% uptime polling 4 crypto exchanges for data. - Hired and managed 3 subcontractors for consulting and outsourcing various parts of work.

- Achieved an above average accuracy at stock market day trading prediction by training logistic regression and Transformer-based classifiers on financial and Twitter data. - Developed the main data-cleaning pipeline for outlier-riddled stock exchange entries.

Took a gap year to study mathematics and machine learning. Solved more than 1500 math problems, proving all my solutions to my mentor. Completed several online courses on machine learning. By the end of the year, figured out most of the foundational principles in ML and implemented key neural network architectures from scratch.

- Launched one of the first decentralized exchange apps by assembling a cross-functional team, making the core architectural decisions, and implementing it in TypeScript and Angular. - Authored a cryptographic API allowing secure client-to-blockchain requests using TypeScript. - Sped up the process of turning mockups into HTML/CSS/JS components more than 4 times. - Designed and implemented a service for aggregation and analysis of blockchain data using Node.js and React for internal usage which allowed the company to sell statistical reports. - Covered the code with Jasmine unit tests, reducing the number of hot fixes by a factor of 5. - Gave talks on specifics of doing frontend for blockchain, including one at HighLoad++ 2017.

- Made a data-streaming service from LabVIEW servers to client applications over WebRTC with Node.js/NW.js for backend and React for the UI which enabled the startup to open sales.

- Drafted the initial architecture for the entire platform and mentored the team on its specifics. - Executed a complex 12 pages web app with data charts using React, Redux, and GraphQL. Unfortunately, the company faced financial constraints, and the development group was dismissed, leading to the cancellation of the project.

- Added features to 3 large projects and translated design mockups into HTML and CSS. - Improved the performance of several frequently used components by a factor of two by leveraging the expertise in jQuery internals and JavaScript profiling and debugging. - Created a Node.js tool improving code quality and increased test coverage from 40 to 85%.

- Completed several small- to medium-sized projects using JavaScript and React for business logic and HTML/CSS for typesetting. Occasionally wrote Python grabber scripts and ETL pipelines.

- Developed a web interface for the queue management system with jQuery, HTML, and CSS. - Planned and put together a map tool for customer companies to track their offices’ current capacity by using OpenLayers map API and jQuery UI. Rolled out for more than 100 offices.

- Led the design and development of a new social network for professional movie makers. - Garnered a strong user base of approximately 250 personal and 30 company accounts.