Philip Shurpik

Experience

• 

  Kwendi

  Nov 2010 - Dec 2012

  Media Market Analyst | Software Engineer

  Started professional journey as an analyst/researcher at a media consulting firm; elevated to a software engineering position within six months.• Engineered a media-audit/media-analysis system with C# and MSSQL, accelerating analysis and presentation preparation by 3-5x.

• 

  Terrasoft

  Jan 2013 - May 2014

  Javascript + C# Engineer

  Developed core features and new modules for bpm'online (now Creatio) CRM• Closely collaborated with business analysts and coordinated three team members

• 

  Krusche & Company

  Sept 2014 - May 2015

  Javascript Developer

  Took part in multiple projects as an JavaScript Engineer• Fully rewrote legacy project using the latest technologies

• 

  EPAM Systems

  Jun 2015 - Feb 2016

  Software Engineer

  Worked in JavaScript team that developed POS terminal for Canadian Tire• Developed AngularJS SPA using TDD and pair programming in a team of 10 engineers• Facilitated communication with business analysts, backend developers, and QAs.• Mentored Epam JS Labs students and engineers in JS Mentoring program

• 

  Ciklum

  Feb 2016 - Nov 2017

  Senior Software Engineer

  Led development of mobile project for Invoicing software company Debitoor• Worked on receipt/invoice OCR implementation using Machine Learning• Authored offline-first mobile app architecture using React Native and Redux• Developed new backend features for the Debitoor web app (NodeJS)

• 

  Self-employed

  Sept 2017 - Jun 2018

  AI Engineer | Data Scientist Consultant

  Was involved in 2 ML projects:• Financial Markets prediction system: data scrapping, statistical data analysis, feature selection and feature engineering, creating machine learning models. • Marketing chat bot with image analysis (facial recognition, hair color detection) - led team of 2 engineers from start to deployment to production. As engineer was mostly focused on CV part and deploying to production.

• 

  Reface

  Jul 2018 - Oct 2023

  Principal Engineer | Head of ML Production

  Started as employee #3 and helped grow the company to 200+ employees and 100M+ appinstalls. Built the core ML pipeline for video face swaps.• Hired, set up and built ML Production / Backend team up to 8 engineers• Built scalable cloud architecture that is able to handle 25+ thousands video swaps perminute and hundreds of thousands API requests per minute• Created architecture of our photo/video processing instances• Worked on research and implementation of the latest open-source models andintegrating them into our pipelines• Initially as ML Engineer took part in development of our first ML models and pipelines(e.g. face tracker) and optimizing performance of neural networks and ML algorithms Show less

• 

  Self-employed

  Sept 2023 - now

  AI Solution Architect | AI Engineer | Consultant

  AI-focused consulting to startups, which includes architectural planning, conductingengineering interviews, coding implementation and deploying to cloud (GCP/AWS/Azure).Key AI projects for startups involved:• LLM voice agent with memory using LangGraph, LLM fine-tuning, RAG with Qdrant DB• ETL pipeline with LLMs and Dagster for AI market research• Solution for furniture object segmentation and item image search using Qdrant vectordatabase for interior design CRM system• Pipeline for scrapping, data enrichment and transforming using LLMs Show less

• 

  Svitla Systems, Inc.

  May 2024 - Feb 2025

  Senior AI Engineer

  Architected and built an LLM-based agent system using LangGraph for Pearl, an AI-powered platform providing personalized mental health support through text and voice interactions.• Implemented application memory architecture using LangGraph for storing and retrieving Core and Recall memories.• Fine-tuned the GPT-4o model to match GPT-4 Turbo’s conversational tone using a curated dataset created by psychologists• Fine-tuned GPT-4o Mini to improve conversation routing accuracy to 95%+• Designed and built a comprehensive data pipeline for audio/video processing, including scraping, transcription, and retrieval-augmented generation (RAG) with Qdrant DB for enhanced content retrieval Show less