Hasnain Akhtar.

Intelligent financial advisory platform for the Pakistani market. ML models for loan eligibility, credit scoring, and financial risk prediction. Leading backend as FYP.

Improved DBSCAN algorithm for detecting small objects (humans, rocks). ~17% accuracy improvement with novel implementation — a first of its kind based on research paper study.

Improved pedestrian crossing prediction on the JAAD dataset. Fused SMPL+CLIP with RGB, optical flow, and pose data — four-component fusion for higher accuracy.

Unsupervised system to detect anomalies in network traffic. Implemented Isolation Forest, LOF, and DBSCAN with feature engineering and PCA/t-SNE visualization.

Intelligent racing agent for TORCS simulator using reinforcement and supervised learning. Neural network with Llama models for decision-making. Q-learning and policy gradient methods.

Comprehensive fuel management system handling inventory, orders, and transactions. Complex stored procedures for automated reporting with role-based access control.