ALEENA KHURSHID

At Corvit Systems, I completed machine learning and deep learning tasks during my internship and built a Roman Urdu sentiment classifier that labels customer messages as negative, neutral, or positive.

For my solo final-year project, VideoGPT, I built a CCTV analysis pipeline for vehicle detection, license plate recognition, and vehicle color detection. I also added natural-language search that turns queries into database queries using a Groq LLM.

I built a Video RAG Assistant that answers questions about video transcripts and cites timestamps. I evaluated retrieval on self-written questions and transcripts, then added pytest coverage and a CI check for retrieval quality.

My Roman Urdu classifier compared a TF-IDF and Logistic Regression baseline with fine-tuned DistilBERT; the simpler baseline performed better. I served the model through FastAPI and added a Gradio demo.