Yurii – Node.js, Nest.js, Typescript

Backend system for doctors to remotely monitor and manage patients using insulin injection devices connected to mobile apps. Continuously collects device telemetry from 1,000+ connected devices, aggregates medical statistics in real time, and surfaces actionable insights on an admin dashboard.

• Architected database schema and REST APIs handling continuous telemetry ingestion from 1,000+ IoT medical devices
• Built aggregation layer processing high-frequency data streams — real-time statistics across insulin dosage, sugar levels, and injection history per patient
• Ensured secure handling of sensitive patient health data with access controls and audit logging
• Collaborated with frontend team on doctor-facing admin panel for patient monitoring

Intelligent customer support automation deployed across Viber and Telegram. Combines a structured FAQ knowledge base for high-frequency requests with dynamic ChatGPT-powered responses for open-ended queries — reducing manual support load without sacrificing response quality.

• Designed two-tier response architecture: rule-based FAQ engine + ChatGPT API fallback for context-aware responses
• Integrated Viber and Telegram Bot APIs with a unified Node.js backend from a single codebase
• Built FAQ knowledge base covering 50+ topics — automated resolution of ~70% of incoming queries
• Optimized response latency under 2 seconds via Python-side caching and async request handling

Two-phase engagement on a law firm billing ecosystem: core SaaS product for billing management, and an integration layer with third-party legal billing systems — RocketMatter, PracticePanther, ActionStep, and others. Detects non-billable work in real time and supports LEDES 1998B import/export.

• Built full backend integration service with NestJS — data normalization across 4+ third-party billing API formats
• Implemented real-time notification system alerting lawyers to non-billable work as it occurs
• Integrated Stripe for payments; Redis caching reduced billing lookup latency by ~40%
• Delivered consistent feature development across a multi-year engagement

Public data analytics platform aggregating and ranking Decentralized Autonomous Organizations across quantitative and qualitative metrics — voting patterns, decentralization scores, member participation, and expansion dynamics. Data aggregated across 200+ DAOs from multiple external sources.

• Owned backend architecture and database design for the full platform
• Implemented Neo4j graph database for DAO relationship modeling — enabling complex multi-hop queries not feasible in a relational DB
• Drove query optimization — reduced p95 dashboard response times by ~60% under concurrent user load
• Set up CI/CD pipelines and owned cloud infrastructure on AWS with Kubernetes orchestration
• Extended platform backend with voting analytics engine, decentralization scoring logic, and member expansion metrics

• Architected backend and database structure for the core ordering platform from the ground up
• Built real-time courier tracking logic with live order status updates visible to customers
• Integrated external restaurant APIs to aggregate food options from multiple providers
• Stepped into Tech Lead role as team scaled to 10 developers — drove architectural decisions and technical reviews

• Developed the the project's architecture and database structure
• Integrated external APIs to aggregate food options from various restaurants.
• Developed courier tracking logic and a user notification system

A New York-based AI platform that analyzes online content at scale to detect misinformation, narrative risks, and coordinated manipulation. Built on deep learning models with a multi-layer verification approach — combining AI-based scoring with human fact-checkers for high-stakes cases. Used by enterprises, media companies, and public sector organisations for brand risk and geopolitical monitoring.

• Developed project architecture and database structure for the full platform from scratch
• Built article credibility scoring system with five-tier rating (unreliable → reliable), sentiment scoring, and publication date filtering
• Implemented search and filtering across sources, keywords, and named entities with word-level breakdown per article
• Developed and maintained all core platform functionality including saved content management and collaborative annotation