Guilherme Forton Viotti
Software Developer Project Manager Sustainability Advocate Show all

Portfolio

image of project 'Temet'

Temet

Know yourself using AI. Allow specialists to help you. Improve your life.
Python Django Machine learning AI CI/CD Large Language Model (LLM) REST API AWS S3 EC2 AI Agent React FastAPI

Temet is your privacy-first AI assistant that lives on your computer. Its main power comes from its ability to read, understand, and help you work with your own documents, securely. Add folders from your computer to give Temet context. Once a folder is "indexed", the AI can answer questions based on its contents, summarize documents, and more. Examples: For a lawyer: Add a folder of laws and contracts and use AI to cross-examine both. For a doctor: Have patients share their medical history folders for a complete, AI searchable overview. For you: Use voice chat to create a daily journal, with each entry saved as a file in a "Diary" folder so you can ask questions about your objectives and milestones. Your data never leaves your computer unless you choose to share it. When you share a folder, Temet creates a direct, encrypted peer-to-peer connection to the other user's app. No central servers are involved in the transfer.

Visit platform open_in_new
image of project 'MIP - Mintec Internal Platform'

MIP - Mintec Internal Platform

Application to manage company back office
Python Django Bootstrap Material Design PostgresQL RabbitMQ Celery AWS MongoDB DRF Docker docker-compose REST API EC2

This was an exceptionally challenging project, only matched by its great success. A platform to deal with the data, manipulation and export of multiple team assets, such as weather data ingestion from API endpoints, automation with Celery and RabbitMQ, MongoDB data read, CSV/Excel easy exports, SSO auth integrated with the company's Azure Tenant and a Machine Learning internal solution that helped the sales team quickly compare all time-series of the company with prospects requirements by context similarity.

image of project 'NFT investment alert system'

NFT investment alert system

Following big money in the Ethereum network
Python Django Heroku

This project was dubbed 'Remora' (fish who swims with sharks and eat small chunks of their prey) with the idea to constantly check for movements of 'sharks' in the NFT market and replicate long or short positions in NFT collections before they would increase or decrease in value. The system sends email alerts every time conditions set by the analyst are met, showing the collections that might become the 'next big thing'.

Etherscan API open_in_new
image of project 'Sci-hive'

Sci-hive

Your scientific content-sharing platform
Python Django Bootstrap AWS Docker docker-compose Nginx SSL Cloudflare CI/CD EC2

This personal project was born from lessons learned on my other sustainability project, IDETRA. The platform allows you to connect data points (posts) to create a storytelling arch with multiple possibilities or assemble them into a network. The references you give for each data point are identified and get a 'sci' tag if they are from a known scientific publisher. The rating mechanism allows the public to score each data point. Uses the Django deployment on AWS featured in this portfolio.

sci-hive.com open_in_new
image of project 'Sarcastic LLaMA3'

Sarcastic LLaMA3

A fine-tuned Model, build for exercise purposes
Machine learning AI Large Language Model (LLM) Ollama HuggingFace Low-Rank Adaptation (LoRA)

Meet Llama3sarcastic, the sassiest AI chatbot fine-tuned to deliver razor-sharp sarcasm with a side of helpfulness. Fine-tuned by me, this model is built on the Llama3 architecture and trained with a curated dataset that ensures every response has just the right amount of snark and wit. Whether you need clever banter, sarcastic life advice, or an AI that "totally loves" answering your most obvious questions, Llama3sarcastic has got you covered. The model is available for download on Hugging Face, making it easy for developers and humour enthusiasts to deploy this sarcasm powerhouse in their projects.

Hugging Face Repo open_in_new
image of project 'LLM fine-tuning boilerplate'

LLM fine-tuning boilerplate

A script to fine-tune a Llama3 model
OpenAI Machine learning AI Large Language Model (LLM) Unsloth Ollama HuggingFace Embedding Google Colab Low-Rank Adaptation (LoRA)

This Google Colab script simplifies fine-tuning Llama 3 models using the LoRA (Low-Rank Adaptation) technique, making it resource-efficient and accessible for creating custom AI models. It integrates seamlessly with Hugging Face Transformers, allowing users to fine-tune specialized models on their datasets with minimal computational requirements. Perfect for anyone looking to unlock Llama 3’s potential with ease.

Google Colab Notebook open_in_new
image of project 'Terra Formus'

Terra Formus

Organizing solutions in an actionable way
Python Django Bootstrap AWS Docker docker-compose Nginx SSL Cloudflare CI/CD REST API EC2

This platform allows people to populate solutions and organize them in strategies that are both actionable and scalable. The proposed informational structure allows solutions to be solid and developed by any willing entities.

terraformus.org open_in_new
image of project 'Plann.us'

Plann.us

Free Project Management Platform
Python Django AWS DRF Docker docker-compose Nginx SSL CI/CD REST API AWS S3 EC2 React

Built to empower individuals and organizations driving positive change, it offers versatile Table, Board, and Roadmap views to suit any workflow. Promote your project on the home page and allow volunteers to pick up tasks.

Visit platform open_in_new
image of project 'Local RAG Agent'

Local RAG Agent

GPT, RAG and Agent modal interface
Python Bootstrap OpenAI Machine learning AI Large Language Model (LLM) Flask Ollama HuggingFace Embedding AI Agent

Explore the power of Retrieval-Augmented Generation (RAG) and agent-based pipelines with this adaptable platform. Your Local RAG Agent GPT supports local and API model execution, integrates tools like Llama Index and LangChain, and offers modules for various workflows. Easily manage documents and databases to create efficient AI-driven solutions tailored to your needs. A work in progress with more features coming soon!

Github open_in_new
image of project 'Multiple Djangos, one AWS EC2'

Multiple Djangos, one AWS EC2

Easily deploy multiple applications under Nginx on different domains (using SSL)
Django AWS Docker docker-compose Nginx SSL Cloudflare

By copying one folder and a Dockerfile on your AWS server, you can deploy multiple Django apps on a free EC2 instance, direct your domains to a free Cloudflare account and have them served behind a Nginx load balancer with the power of docker-compose. This is not an automated solution by default, as it is designed to be built upon. Just follow the instructions and use the suggested 'quick Django deploy' image for a highly effective and lighting quick 'develop and deploy' combo.

Github open_in_new
image of project 'RAG GPT'

RAG GPT

It's ChatGPT on steroids - You upload a document and instantly augment it with your data!
Python AWS OpenAI Machine learning AI CI/CD Large Language Model (LLM) REST API AWS CDK AWS Lambda AWS S3 Flask

The application is an advanced version of ChatGPT that allows users to upload a document and augment ChatGPT's responses with data from their uploaded file. It is a prototype designed to showcase CI/CD Github actions pipeline, AWS services (CDK, Lambda running Flask, and S3) and the capabilities of Large Language Models (LLM) using the Retrieval-Augmented Generation (RAG) technique.

Visit Github repo here open_in_new
image of project 'Django quick deployment'

Django quick deployment

Run one single command and have an operational Django app locally, ready to be worked on
Django Docker docker-compose

This derivative work from a friend allows you to install a Django app in seconds, with git and docker implemented out-of-the-box, ready for you to develop your application. I aggregated the docker-compose component that allows you to deploy on AWS seamlessly among other details. This is my way to give back to the community that helped me so much over the years.

Github open_in_new
image of project 'Twitch Rewards'

Twitch Rewards

Rewards club for twitch.tv streamers
Python Django Heroku

Platform developed to give the company a competitive edge over all players in the twitch streamers market. Simply put, collect points the more you watch a company partner streamer on twitch and later exchange your points for prizes. The biggest advantage was give users the possibility to change their usernames on twitch and not lose acquired points on our platform - which in the competitor platform was impossible to do.

Twitch.tv API open_in_new
image of project 'This portfolio'

This portfolio

A web portfolio build from the ground up
Python Django Bootstrap Material Design AWS Docker docker-compose EC2

To showcase what I can do as a software developer, I build the portfolio web platform that you are navigating right now from scratch. From comments to clean code, tests to quick deployment, it's all here. Click on the button below to see the source code on Github.

Github open_in_new
image of project 'IDETRA'

IDETRA

Education and project management platform
Python Django Bootstrap AWS Docker docker-compose EC2

This is a personal project of mine. I created this platform to absorb volunteers willing to work on a humanitarian cause. From the start it was a very complex platform that encompass explaining what the initiative is all about, an education platform and a project management system. Completely build from scratch, I coded all the frontend and most of the core backend, eventually needing help to finish some of the more complex functionalities since it was my first Django project.

idetra.org open_in_new