I'm a Computer Science graduate from Sukkur IBA University (2022β2026), based in Pakistan, building research-grounded AI systems β Vision-Language Models, RAG pipelines, speech recognition, and computer vision β backed by rigorous benchmarking and evaluation.
- π Currently building AI/ML systems spanning VLMs, RAG, and speech AI
- π Accepted Paper Presenter @ iCOMET 2026 β Localized Benchmark Evaluation of Vision-Language Models for Visually Impaired Assistance (IEEE Xplore, expected)
- πΌ Former AI/ML Developer Intern @ Pakistan National Assembly Secretariat
- π₯ 1st Prize β SIBA Fest 2024 Software Project Exhibition
- π± Open to AI/ML Engineer, Deep Learning Engineer, and Research Assistant roles
| Project | Description | Stack |
|---|---|---|
| SceneNarrator β VLM Benchmarking & Fine-Tuning | Assistive scene-narration & navigation system for the visually impaired; LoRA fine-tuned LLaVA-7B, benchmarked against CogVLM/BLIP-2. Accepted @ iCOMET 2026, IEEE Xplore-bound. | LLaVA PyTorch PEFT LoRA |
| Personalized RAG Chatbot | Document ingestion β embeddings β semantic retrieval β grounded generation for knowledge-aware Q&A. | Python RAG Chroma DB |
| Speaker Recognition System | Single/multi-speaker recognition pipeline: feature extraction, classifiers, validation & evaluation reporting. | Python Audio Processing |
| SAM3 vs Human Segmentation Validation | Human-in-the-loop tool comparing SAM3 masks against ground-truth annotations via IoU, precision, recall, F1. | PyTorch SAM3 Gradio |
| Automated Toll Tax / Smart Toll Vision System | Real-time vehicle detection, tracking & counting for dynamic toll estimation with dashboard overlays. | YOLOv8 DeepSORT/ByteTrack OpenCV |
| XAI: Virgin vs Recycled Plastic Classification | Swin Transformer classifier with Attention Rollout explainability and confidence thresholding. | PyTorch Swin Transformer |
| Fine-Grained Emotion Classifier | Compared classical ML, BERT & RoBERTa on Ekman-emotion data; fine-tuned RoBERTa to 89.0 F1 / 89.4% accuracy. | Transformers RoBERTa BERT |
| Anemia & Leukemia Disease Prediction | Dual-disease healthcare prediction pipeline with SMOTE balancing and gradient boosting. | XGBoost LightGBM |
| Bank Management System | Secure desktop banking app with SHA-256 transaction handling and role-based admin controls. | Java MySQL |
See GitHub repositories for code.
BS in Computer Science β Sukkur IBA University Β 2022 β 2026 Coursework: AI, Machine Learning, Deep Learning, Computer Vision, DBMS, Software Engineering, OOP, Web Development, Cloud & Parallel Computing.
- π iCOMET 2026 β Accepted Paper Presenter, Localized Benchmark Evaluation of Vision-Language Models for Visually Impaired Assistance (IEEE Xplore, expected)
- ποΈ Google AI Essentials Β· Google Prompting Essentials Β· Google AI Specialization Β· AI for Research and Insights (badges on Credly)
- π₯ 1st Prize β SIBA Fest 2024 Software Project Exhibition
- π Top-10 β Grand Tech Expo (Bank Management System)
- ποΈ Special Mention Award (Γ2) β Model United Nations, representing the People's Republic of China