Build autonomous Agents like Legos

Develop and use pre-build code blocks like Legos to build arbitrary powerful autonomous agents in minutes.

I initiated this project myself. I collaborated and consulted with startups to build and discuss AI-based autonomous agents. I identified several components necessary for each agent to construct an autonomous agent and decided to simplify the entire process with Eezo.

Visit https://eezo.ai (legacy website, new website is launching soon)

The realization of this project involved speaking with users, conceptualizing a solution based on their needs, and designing the UI/UX in Figma, along with tests which included a chat interface. I built the frontend, incorporating a Slack-like UI block system that can be dynamically generated by AI and assembled in the backend using Vue.js and TypeScript. Planning and constructing the backend relied on Google Cloud + Firebase, Supabase (PostgreSQL), utilizing Python and TypeScript. I developed a WebSocket-based RESTful API that enables users to connect their code to the AI, employing a variety of prompting techniques and advanced RAG, among other methods, around OpenAI's LLMs. I designed and coded the website to showcase real examples of the product's capabilities. For organization and brainstorming, Notion and Figma were my primary tools.

OneReview is a product developed by our team consisting of Harry, Eliska, and Ben. 

We won first prize for it at the Hackathon held at On Deck in San Francisco. OneReview is a Chrome beta extension that summarizes the most important customer reviews into one trusted review. It saves time by providing concise highlights and pros/cons of products, eliminating the need to read through multiple reviews.

It detects real reviews and filters out fakes, ensuring reliability. Users can also compare products and their highlighted reviews side by side.

The Chrome extension is built with TypeScript and Vue. The server is built with Python, scraping information from Amazon using ScraperAPI and analyzing reviews in an LLM pipeline centered around Anthropic's Claude LLM.

More infos: https://yep.so/p/onereview

ROKIN is an engineering-focused search engine that improves search efficiency and collaboration. It sends tailored queries to top search engines, filters results, and organizes them for engineers. Teams can collaborate, share relevant content, and avoid duplicating work.

This project was supported by Fraunhofer Ventures, Esa, Media Lab Bayern, and the SCE. 

My task involved building the entire frontend, which included features such as managing search result lists, collaboration tools for teams to jointly search for parts for large projects with real-time chat, and a marketplace to share and sell existing lists of well-curated search results for highly specialized projects.