Rodrigo Heck

A system that can show the brand prevalence in an image of a supermarket shelf.

I developed the AI algorithms necessary to segment and classify each product and the platform that would enable the client to manually label the brands for training purposes. All solutions were encapsulated in a simple-to-use API that received an image and returned the distribution of products and brands.

An R Shiny report that allowed our clients to interact with their summarized data.

Our company collected a lot of social media comments about our clients. Still, most of the interaction they had with this data was by checking the overall sentiment displayed as a single number.

My intention with this project was to automatically generate a Shiny application that offered all the important information they wished to know in a single place. They had access to previously calculated metrics and could interact with the graphs and comments, conduct customized searches, and generally have a better overall understanding of their data.

The clients to which we introduced this solution were satisfied with the tool. They enjoyed this guided but not static experience to know what happened over a certain period regarding their social media presence.

A platform that connects consultants and clients. I was the CTO and developed all the functionalities available, combining NLP techniques for the search algorithm and front and back-end skills to deploy an easy-to-use layout. Consultants have the ability to receive proposals to connect, they can create live events and sell tickets, and they can offer their skills to specific companies. These features required the correct implementation of a flexible payment gateway. In this case, I used Stripe. The development of an online conference solution to in-platform video connection; the use of front and back-end frameworks to build everything up (Flask, Firebase, and React), and the deployment of the system on cloud servers (AWS).

A Shiny app able to access different types of datasets and convert all of them into a unified format. Surveys conducted by a research team were pretty similar, but there were some idiosyncrasies that made the merge between different datasets somewhat complicated. This app allowed the team to map all the questions into the same structure and download a single merged dataset.

This was a project for a law firm that wanted to enhance their internal knowledge database, hoping to offer a similar experience to what we all enjoy with ChatGPT. First, it required building a retrieval model to get the relevant documents, then tweaking GPT-4 enough for it to understand what it is supposed to do with its context window, and, finally, building a compelling and beautiful interface for users to interact.