Oleh – Python, Django, SQL

AI-powered market intelligence platform: AI agents swarm, news monitoring, multifactor risk models, automated ETL, and backtesting infrastructure

• Led development of a comprehensive market intelligence platform for a hedge fund.
• Built autonomous AI agents for real-time news monitoring and analysis.
• Designed AI-based ETL workflows on AWS Step Functions for continuous data ingestion.
• Engineered infrastructure for calculating and running weighted multifactor risk models across all economic sectors, using AI coding tools to automate model transformation and execution.
• Implemented offline QuantConnect for backtesting and future live-trading automation.
• Delivered a suite of data mining tools used by risk managers for designing new models and generating insights about companies and sectors.
• Built FastAPI services with PostgreSQL-backed long-term memory, chat, and feedback loops.
• Refactored agent architecture to MCP-based tool routing.

Jupyter-to-production GenAI pipeline for a global rail operator

• Served as middleware engineer transforming Jupyter notebook AI prototypes into production-grade FastAPI microservices for a global rail operator.
• Designed end-to-end pipelines for ingesting operational data, integrated React frontend components for AI-powered insights, and ensured secure, low-latency workflows for live stakeholder demos.
• Bridged the gap between data science experimentation and production deployment.

Retrieval-augmented chatbot grounded in a digital agency's knowledge base

• Architected and delivered a Retrieval-Augmented Generation chatbot for a digital agency's website.
• Implemented document chunking, embedding generation, vector storage, and context-aware retrieval to provide accurate, grounded responses from the agency's knowledge base.
• Designed the conversational flow and integrated the chatbot into the existing web frontend.

Embedding-powered Classifiers for candidate screening that cut human review time by 65%. The resulting system screened IT candidates end-to-end from resume ingestion through AI-driven scoring.

• Joined the founder to evolve a prompt-based PoC into a functional MVP.
• Redesigned the screening pipeline around OpenAI embeddings and custom classification models, replacing brittle prompt-only logic with a structured FastAPI service.
• Integrated with the React frontend for real-time candidate dashboards.

CPU-based sparse inference platform and open-source vLLM contributions

• Contributed to a large-scale AI model optimization platform as part of an established product team.
• Developed and maintained FastAPI endpoints for CPU-based sparse inference.
• Integrated the REST API layer with the React frontend and the DeepSparse engine.
• Contributed to the open-source vLLM repository.

Internal ad campaign analytics platform with dashboards, team collaboration, and access management for 50+ music labels

• Developed the backend for a Flask/PostgreSQL/Snowflake analytics platform serving the internal music label activities.
• Integrated and transformed terabytes of campaign data for real-time reporting and team collaboration.
• Optimized query performance for peak loads during major releases, and designed dashboard features.

Web application for legal billing and a vendor management platform

• Performed backend upgrades for an enterprise legal billing and vendor management platform under rapid growth.
• Worked with Django, Redis, and PostgreSQL to maintain 99.99% uptime.
• Delivered features across billing workflows, vendor onboarding, and reporting.

A multi-platform secure messaging system for healthcare providers (HIPAA Compliance)

• Resolved performance bottlenecks, ensured HIPAA compliance across the stack, and maintained legacy C++/Python integrations.
• Focused on reliability and security for a mission-critical communication platform.

BI dashboards, custom data labeling tool, and CNN-based product image classification via transfer learning for retail operations

• Built end-to-end retail analytics pipelines and BI modules powering operational dashboards for retail clients.
• Developed a custom data labeling tool used to accumulate training data for CNN transfer learning on product image classification.
• Built analytical reporting features using Django/DRF.
• Implemented security features including two-factor authentication.