Built Swift Bird Chatbot with RAG and prompt engineering on Mistral Large 2:
- Implemented voice mode with Text-to-Speech and Speech-to-Text
- Added image comprehension with Pixtral
- Support for multiple languages 🇬🇧, 🇫🇷, and 🇪🇸.
As part of my master research internship I worked on research topic "Exploring Image Representations of Time-Series for Enhanced Model Performance". The significance of my work is in synthetic dataset generation for large-scale time-series data.
I've also had the opportunity to contribute to the following projects:
This project compares the performance of LLMs (DistilBERT and BERTweet) to conventional models such as CNN in two tasks:
- Classification of Emotions
- Classification of Sentiments
An Android project that uses OpenCV Android to detect intruder in low-light conditions, taking the snapshots with flashlight, uploads it to the cloud and notifies the owner.
AgriFarm won best app of the year award in Google Android Developers Challenge for its innovative approach to detecting crop disease with on-device Machine Learning.
Uses offline machine learning to detect thumbs-up/thumbs-down gestures as an interactive approach to collect user feedback for surveys. Offered as an Android app.
The Python version of SOMToolbox
Deep Embedded Self-Organizing Map: Joint Representation Learning and Self-Organization
SOMperf: Self-organizing maps performance metrics and quality indices