Antoniu – Typescript, React, Node.js

A platform where marketers can create ad-ready content following strict brand guidelines and publish it on multiple advertising platforms (Meta, TikTok, Google Ads, LinkedIn, etc)

• Designed and shipped an AI agentic system using Mastra (LLM orchestration) that analyzes user data and competitor PPC campaigns to recommend custom ad creatives for paid platforms.
• Developed publishing pipelines integrating with Meta, Google, LinkedIn, TikTok, and Amazon ad platforms, including platform-specific creative validation and pre-submission rule enforcement.

An AI-powered platform for recommending gifts.

• olo developed an AI-enhanced recommendation platform that was giving gift suggestions based on a set of details about the giftee
• Hosted the app on a VPS and managed it through Plesk (which also provided continuous deployment)
• Used ChromaDB for vector storing of the gifts, NextJS v15 for Frontend, Flask for Backend, Grafana for Observability
• Used Claude agents + playwright for automated scraping of websites

In the early era of AI, the Create module was designed to provide an AI-first interface that would let the platform's users generate images from text, similar images, and strict brand guidelines.

• Created a new and totally untested feature and a testing plan for the existing create module feature, both unit testing (Jest) and e2e testing (Playwright)
• Provided the initial set-up for the testing environments and settled on the patterns of testing and code standards, and led the initiatives until a high coverage/confidence number was reached
• Developed draft functionality using Jotai for state management
• Delivered under critical time, the ability to create images based on a given set of reference images
• Put the foundations of the first MVP with a first version of 4 generated images based on text and guidelines
• Built a predictive prompt-recommendation feature on top of the agentic pipeline, reducing user abandonment in the creative-creation flow by ~30%. Used Mastra AI, Vector Databases, and RAG

An enterprise platform used to create, manage, and deliver digital content, websites, mobile apps, and user experiences across multiple channels. It is acting as both a DAM and a CMS.

• Led a major frontend refactor that made the codebase reusable across resource types, cutting new micro-frontend (MFE) delivery from months to ~3 sprints. Worked with React, TypeScript, Adobe Spectrum, Redux, Redux-Saga, Monorepo environment, Lerna
• Developed features such as async polling of tasks, unified error handling, retry logic mechanisms, content editing, displaying paginated content in an infinite scroll view, filtering
• Developed the contextual search through content resources capability using RAG [Hackathon]
• Integrated other internal frontend applications and managed communication between micro-frontends through custom events.