Kibrom Kidane

Driven by a complex algorithm in the back end, Dine is a simple-to-use food delivery app I developed for a startup company in Addis Ababa. It features location-based filtered restaurants and a complete food ordering and tracking system.

Flutter-based audio streaming and podcast app used by thousands of users. I was a Flutter developer primarily focused on the project's back end and bloc-related parts. I collaborated with other developers to implement and improve the code's data model, data flow, and efficiency.

Neuron is an airdrop web application designed with React and is intended to function as a Telegram mini-app. This tap-to-earn project allows users to tap a coin to receive an airdrop once the Neuron coin is listed.

This project focuses on enhancing an open-source LLM like Llama 2's ability to detect and understand logical fallacies, a crucial aspect of critical thinking and argument analysis. The work involved creating an extensive dataset of logical fallacies and fine-tuning large language models for improved fallacy recognition.

KEY ACHIEVEMENTS

1. Dataset Development:
• Curated organic datasets from reputable sources, covering 14 fallacy categories.
• Created a synthetic data generation pipeline, expanding the dataset to over 10,000 examples per category.
• Ensured diversity across multiple domains while maintaining consistency in fallacy representation.

2. Model Fine-tuning:
• Utilized the Anyscale platform to fine-tune Llama 2 and 3 models on the custom dataset.
• Implemented systematic approaches for model training and evaluation.

3. Tool Development:
• Developed Python scripts for data generation, validation, and model testing.

RESULTS AND IMPACT
The project yielded significantly improved accuracy in fallacy detection and generation. Potential applications include educational tools, content moderation, and argument analysis systems.

I developed a Python project to manage trading accounts and equity. I also created an algorithm to decide trade probability and to help with position sizing. Finally, I integrated the service with other APIs to get financial information.

A 2D canvas app that children play to create different shapes.

I created 2D canvas shapes that can connect and rotate. I also optimized canvas rendering and drawing alignment issues. Finally, I built the app with React and a Vanilla JS canvas library.