Vasyl Boyko

Contributed to a large-scale food delivery courier app serving over 300,000 users across Europe and Africa. Worked on enhancing the existing app built with React Native, Redux, RxJS, and TypeScript. Delivered several mid-to-large-scale features, including the development of multiple screens and a dynamic multi-screen form configured from the back end. Initiated and implemented end-to-end testing with Detox, enabling the creation of a robust test harness. Developed several Node.js microservices with endpoints for the app and created React-based pages in the admin panel to streamline business process management and configuration of settings.

Rescued and launched a Next.js web app after an incomplete handoff from a freelance developer.

I inherited a partially built Next.js app with messy code, tech debt, and no documentation. After investigating the codebase, I fixed critical bugs, integrated it with the back end, and implemented a complete redesign based on Figma.

I successfully delivered a stable, production-ready site within the expected timeframe.

Viv Technologies Connected Home Care is transforming the home care industry through cutting-edge digital solutions, including software-as-a-service (SaaS), platform-as-a-service (PaaS), and mobile applications. These tools are purpose-built to streamline operations for multi-site home care businesses, enhancing efficiency in telehealth, home care management, and business processes.

Key contributions:
• Developed a 3rd-party data import tool to enable seamless transitions from other home care management platforms, transferring existing client, caregiver, and visit information into Viv Technologies Connected Home Care.
• Created new web pages, endpoints, data models, and business logic to support this feature and improve platform functionality.

WordProof is a groundbreaking platform revolutionizing trust and content protection. Tailored for publishers, content creators, and eCommerce businesses, it offers timestamping solutions to validate content integrity. Featuring copyright protection, time machine, and verifiable trust, WordProof enhances transparency by openly showcasing content changes. The inclusion of structured data for SEO ensures authenticity is verifiable for users, search engines, and social media. With over 5.8 million timestamped items, WordProof helps users secure digital assets and prepare for the future of content management.

In my role, I contributed by implementing half a dozen admin panel pages based on custom Figma designs using Vue. Additionally, I created a UI kit of web components using Stencil.js, designed to be utilized across various company projects. Furthermore, I developed a framework-agnostic web component that provides a timestamping and verification blockchain solution, intended to be seamlessly embedded into customers' websites using Stencil.js.

Obsidian is a business service intelligence and analytics platform situated within the domain-agnostic AIOps market. Its primary objective is to facilitate the alignment of IT services with business goals. The platform offers real-time, historical, and predictive metrics and indicators presented through innovative dashboards and reporting.

My contribution to the project involved developing features for the MVP version of the user interface, specifically focusing on architectural modeling of IT services. The developed UI facilitated capturing their dependencies on both internal and external services, as well as assessing their impact on business processes.

Led the development of a national platform to track and automate continuous professional development (CPD) compliance for lawyers across Ukraine, in accordance with the Ukrainian National Bar Association (UNBA) standards. The platform integrates directly with official UNBA lawyer profiles, providing real-time tracking and certification of CPD activities.

Played a key role in engineering business process automation, designing BPMN flow diagrams, defining system requirements, and wireframing the user experience. Acted as a bridge between business stakeholders and the development team, while also contributing hands-on as a JavaScript front-end developer, specifically in migrating application views from Laravel Blade templates to a modern Vue.js-based front end.

Key contributions:
• Architected automation workflows for CPD tracking and certification.
• Created BPMN diagrams to model complex business processes and ensure stakeholder alignment.
• Defined technical and functional requirements, and produced wireframes for UI/UX.
• Integrated the platform with the official UNBA lawyer profile system.
• Led a development team and coordinated design and engineering efforts.

I architected and built a zero-to-one mobile application prototype for a next-generation dating/matching service powered entirely by conversational AI and vector similarity. As the sole full-stack developer, I built the entire React Native and Expo front end from scratch. On the back end, I designed the core AI orchestration logic using LangGraph to create a stateful, conversational agent that dynamically profiled users through a natural, witty chat interface. Furthermore, I engineered the matching engine by translating these user profiles into vector embeddings and utilizing PostgreSQL (pgvector) to perform highly performant, sub-second matchmaking queries. This project successfully proved out the technical viability of combining dynamic LLM orchestration with vector-based database matching in a production-ready mobile architecture.

I engineered the back-end architecture and spearheaded front-end modernization for a large-scale B2B eCommerce platform. On the back end, I designed and implemented new API architectures and complex MongoDB/PostgreSQL schemas to support a seamless user onboarding checklist. To resolve significant front-end technical debt, I architected a micro-frontend (MFE) solution, embedding a modern React 19 application within a legacy React 16 host. This unlocked the adoption of next-generation tooling like Rspack and Biome. Finally, I pioneered an AI-driven, self-service data ingestion pipeline utilizing an LLM inference engine. This engine automatically parsed and mapped arbitrary unstructured customer data (CSV/XLSX files) into rigid corporate schemas, slashing the customer onboarding timeline from several weeks to under an hour.

I architected and deployed a read-only SQL Model Context Protocol (MCP) server that acts as a secure bridge between production databases and AI agents. The core challenge was enabling non-technical stakeholders (e.g., a CEO) to safely query and explore data without risking data exposure or corruption. To solve this, I engineered a "Zero-Trust" privacy layer using dynamic configuration via Zod schemas to automatically mask personally identifiable information (PII) before it enters the LLM context window. Additionally, I implemented intelligent data truncation as a distinct architectural concern to prevent LLM context overflow during large SQL data retrievals. This project eliminated a high-friction engineering bottleneck by empowering executives to autonomously build operational reports from production data.

I architected an end-to-end automated pipeline leveraging Google NotebookLM as the core intelligence engine to eliminate "post-webinar content debt" for an organization. I built programmatic access to NotebookLM via an open-source CLI, bypassing the lack of an official API to quickly deliver business value. I then engineered a "Human-in-the-Loop" workflow by funneling AI-generated blog and social content into GitHub Pull Requests. This successfully onboarded non-technical marketing staff into a high-integrity, version-controlled developer flow for content review. The resulting "Digital Assembly Line" automatically pushed finalized content from merged PRs directly to social channels and company blogs, significantly reducing manual marketing overhead.

Initially conceived to index 132 million open government court verdicts, I pivoted this project into a "Dynamic RAG" legal preparation pipeline when legacy government servers proved too fragile for mass web crawling. When mass crawling risked DDoSing government infrastructure, I engineered a targeted, agent-orchestrated data extraction flow where an LLM agent brainstormed search heuristics to fetch focused sub-corpuses of highly relevant verdicts. I built a summarization and embedding engine to process these dense legal documents into a searchable vector database. I then developed an MCP server to expose this high-signal vector database to a Legal Assistant Agent for deep semantic search and case drafting. I successfully validated the architecture by dogfooding the pipeline to solve a complex personal legal hurdle involving international diploma acknowledgment.