At Mofid Securities, the largest privately-held brokerage in Iran, I transformed AI concepts into practical products. For instance:Built a conversational AI assistant with OpenAI's API to provide personalized financial guidance to customersImplemented speech recognition to automatically transcribe videos into text summaries, extracting key detailsServed some Open-source projects in production such as PyAnnote Diarization and Whisper for local usageTrained Whisper for speech to text using local data and YouTube crawled videosUsing Prompt Engineering approaches to use LLMs such as OpenAI and Claude effectively

I worked under the supervision of Prof. Hamid R. Rabiee on these tasks:- Developed and implemented a novel approach to enhance YOLO algorithm's performance in detecting polyps, utilizing semi-supervised learning and interpretability.- Integrated pseudo-labeling for semi-supervised learning, improving algorithmic performance on both labeled and unlabeled data.- Adapted PyTorch-GradCAM library to work with YOLOv5, enhancing transparency and interpretability in real-time polyp detection.- Customized YOLOv5 for real-time detection of brain aneurysms in 2.5D CT scans.- Modified YOLOv5 network structure to process multiple images as a single input, optimizing aneurysm detection accuracy.

The projects I worked on during my four years with Sepand Inc. were as follows:- Designed and implemented RESTful APIs using C# and .NET Framework to be used in Android ClientQuery Optimization in SQL-Server- Report generator for DataGridView in StimulSoft Report- Integrating different devices (POS, Scale, SMS, TelegramBot) with Sepand Application- Database Schema Comparer and Upgrader using DACPAC

I worked on the development of a Windows-forms application for IranKhodro. This application focused on managing parking lots, controlling vehicle entrance and exit, and generating management reports. It marked my initial exposure to programming languages and provided hands-on experience in crafting solutions for real-world challenges.