Michael Muzafarov Email & Phone Number

New York, New York, United States

New York, United States

Developed Asset Classification Machine Learning Model:-Classified over $570 billion in client assets with 97% precision -Resulted in a 63% reduction in total number of unclassified assets across entire Addepar Platform-Saved analysts 15,000 hours of otherwise manual work-Improved client experience, allowed for better use of existing platform features, and spearheaded development of new products-Developed with PyTorch Lightning and run on Databricks platform-Used MLflow for logging and model versioning-Implemented Hugging Face’s pretrained Bert LLM as backbone of PyTorch neural networkPioneered First ML Infrastructure at Addepar:-Designed staging-to-production pipelines and documented best practices-Created customizable ML library-Can be leveraged to make new ML models without starting from scratch-Allows any developer to take advantage of out-of-the-box NLP text classification, regardless of ML knowledgeBackend Development:-Utilized PySpark to implement end-to-end ETL pipelines and data quality checks-Integrated new micro-services with existing distributed system architecture -Wrote extensive unit tests for all code contributions with the Python Mock Object Library

New York, United States

Developed Full Stack App:-Allowed salespeople to connect with clients more frequently and efficiently-Fostered an increase in revenue and more results from sales related spending-Python Flask, React, PostgreSQL, AWS, Kubernetes, Jumpcloud SSO SAML 2.0 integrationCreated Data Collection Pipeline for ML Model:-Multi-threaded aggregator collects raw data into an S3 bucket, which triggers a Lambda function to clean and store it in DocumentDB-Pipeline retrieved and stored over 100,000 data points used to train a machine learning model -Model used in new analysis products that provide clients with actionable investment insights