AI/ML & Python Developer
I build production-grade AI systems — most recently a full-stack RAG pipeline (FAISS + OpenAI + FastAPI) that answers questions strictly from source documents, with zero hallucination beyond the retrieved context.
I'm a Computer Science undergraduate at SZABIST Islamabad (Class of 2027), focused on building AI-integrated systems and Python tools that work end-to-end, not just in a notebook.
My most recent project, PDF Oracle, is a full-stack Retrieval-Augmented Generation system: a FastAPI backend that extracts and chunks PDF text, embeds it with Sentence-Transformers, stores it in a FAISS vector index, and answers queries strictly from the top-3 matched chunks via OpenAI's gpt-4o-mini — no hallucinated answers from outside the document.
I've also led a 4–6 person student team through two semester-long software projects, from design to delivery, coordinating module integration and reviewing teammates' code before it shipped.
A full-stack Retrieval-Augmented Generation (RAG) system that turns any PDF into a queryable knowledge base.
Challenge: Preventing hallucinated answers by constraining generation strictly to the top-3 retrieved chunks instead of open-ended LLM knowledge.
Impact: Reduced manual document lookup to a single natural-language query.
Exploratory data analysis on a multi-thousand-row Netflix dataset.
Challenge: Cleaning and reshaping inconsistent multi-value fields (cast, genres, countries) into analyzable structures.
Impact: Demonstrates end-to-end data wrangling and visualization without a framework doing the work for you.
A multi-class banking application with authentication and transaction flows.
Challenge: Designing class boundaries (accounts, transactions, auth) that stay maintainable as features grow.
Impact: Demonstrates OOP fundamentals applied to a real stateful system.
Want to talk shop, collaborate, or about an internship? Reach out.