Petur Subev

A web-based dashboard page for creating multi-message campaigns. Messages can have different audiences or content based on user attributes. The whole layout is organized in a tree-like fashion. Built with Vue.js and TypeScript.

An analytics web page that provides a visual representation of where the users 'went through' in a given campaign. An interactive tree shows is rendered where the width of the branches is calculated based on the users' count. Similar branching is rendered to represent the conversion of users.

A web-based page that allows clients to browse, search, and export all campaigns. The layout is organized in a table-like fashion. The analytics and campaigns come from different data sources, which made it quite challenging.

A web-based interactive source editor similar to JS Bin that allows you to share small examples and run them. It was tightly integrated into the Kendo UI Demos page, that way you don't have to start from scratch but rather, use an existing demo.

A small date range picker that is ready to be copy-pasted into your project. Created in a couple of hours after a discussion of "how hard it is to write components in-house?" The same could be said for tables and dropdowns.

At V7, I led the front-end development of the AI Workflow Builder—an advanced visual interface that enables users to design, configure, and deploy complex ML pipelines using a node-based system. I was responsible for architecting and implementing the entire UI layer using Vue.js and TypeScript, ensuring modularity, responsiveness, and performance at scale.

The builder allows users to chain actions such as model inference, human review, logic branching, and QA into visual workflows tailored to each dataset. I collaborated closely with product managers and back-end engineers to translate dynamic graph logic into an intuitive, drag-and-drop interface.

To maintain high responsiveness while handling large graphs, I implemented virtual scrolling and smart rendering techniques. I also designed reusable components and internal tooling that significantly reduced development time for new workflow features. This builder became a core part of the V7 product offering, improving customer onboarding and automation speed across enterprise clients in biotech, autonomous driving, and retail.

At V7, I was a core contributor to V7 Go, focusing on the design and development of the Table View and Entity Page—the central interface for managing structured datasets and AI-ready file metadata. This interface allowed users to define custom fields, view and sort annotated data, and interact with ML models through a flexible, spreadsheet-like UI.

I implemented the Table View using Vue.js and TypeScript, architecting it for performance, scalability, and extensibility. I introduced virtualization and lazy loading to handle tens of thousands of data rows with smooth scrolling and interaction. The data model supported custom field types (text, dropdowns, files, booleans, etc.), all dynamically configurable via the UI.

In close collaboration with design and product teams, I helped shape an experience where users could define, annotate, and track metadata across large-scale AI pipelines—powering tasks like multi-label classification, data filtering, and agent-driven automation.

The result was a flexible and intuitive workspace that served as the primary control panel for V7 Go customers, accelerating data readiness for training and deployment.