Experience
Built by shipping, not just studying
An AI/ML internship at NCAI, six independent projects, and two published research papers.
AI/ML Research Intern
National Center of Artificial Intelligence (NCAI), UET Lahore · Lahore, Pakistan
- Investigated AI security, LLM hallucination, and reliability through literature review, experimentation, and evaluation of modern AI systems.
- Researched methods for hallucination detection and mitigation in LLM-based applications.
- Currently developing a research study on malignant melanoma detection from H&E-stained histopathological images using deep learning.
- Exploring semi-supervised learning, self-supervised learning, and representation learning to improve medical image analysis with limited labeled data.
- Conducted dataset and literature analysis to identify suitable benchmarks and establish an experimental research pipeline.
Independent Full-Stack & AI Projects
Self-directed · Lahore, Pakistan
- Designed and shipped 6 independent full-stack and AI/ML projects end-to-end — from data model to deployment-ready build — covering web platforms, an Android app, a research pipeline, and a database engine built from scratch.
- Took every project through the same discipline as production work: architecture decisions up front, incremental builds, and a documented account of trade-offs and lessons for each one.
- Worked across the full stack repeatedly: React/Next.js front ends, Node/Flask/FastAPI backends, MongoDB/PostgreSQL/Supabase data layers, and JWT-based auth.
Independent Researcher — MedInsight
Self-directed · Lahore, Pakistan
- Authored "MedInsight: Evaluating Retrieval-Augmented Vision-Language Models for Evidence-Grounded Medical Image Understanding," DOI-registered on Zenodo.
- Built a CLIP-embedded FAISS retrieval index over ~90K ROCOv2 radiology image-caption pairs and used it to evaluate a BLIP-2 vision-language model with and without retrieved evidence on the VQA-RAD benchmark.
- Ran paired bootstrap significance testing rather than reporting a raw accuracy delta, and reported the result honestly — a consistent but not statistically significant effect — instead of overstating it.
Undergraduate Researcher
UET Lahore · Lahore, Pakistan
- Authored an independent research paper "AI-Based Depression Detection from Social Media Using LSTM Networks," DOI-registered on Zenodo.
- Designed the full research pipeline — preprocessing, sequence modeling, evaluation — and documented methodology and findings to publication standard.
Recognition