1. AssetSentinel
A comprehensive solution for monitoring electrical assets in complex industrial networks. It combines real-time telemetry processing with ML-based decision intelligence to provide early warning systems and predictive degradation analysis.
Key Features: Network topology management, Anomaly Detection using statistical learning (z-score standardization), State Classification (NORMAL/WARNING/CRITICAL), temporal consistency, and interactive KPI dashboards.
FastAPINext.jsPostgreSQLMachine LearningPyTorch
2. Multi-Angle Identity Fusion System
An advanced AI-powered surveillance and tracking system designed to maintain persistent identity recognition of vehicles and individuals across multiple camera angles. By integrating datasets such as VeRi-776, CityFlow, and VehicleID, the model learns robust appearance signatures for identity continuity.
Primary Goal: Build a scalable cross-view intelligence system for smart surveillance, traffic analytics, and autonomous city infrastructure, leveraging multi-view computer vision, feature embedding, and re-identification (ReID).
Status: In development / Future Project
3. ORCA EYE
An intelligent real-time assistive navigation system designed to enhance mobility and situational awareness for visually impaired individuals. Built on a hybrid architecture of edge AI and server-side processing, it integrates live camera input, environmental mapping, and spatial analysis to provide low-latency audio feedback.
Core Focus: Combines computer vision, contextual scene understanding, and path prediction to safely identify free paths, moving obstacles, and human presence.
Status: Upcoming Project
4. A* Search Systems Study: Heuristic Design, Optimization & Analysis
A research-grade implementation and empirical evaluation of classical informed search algorithms (A*, Weighted A*, Greedy Best-First Search, Uniform Cost Search, and IDA*) applied to the n-puzzle domain.
Research Focus: Comprehensive evaluation of heuristic design (Manhattan Distance vs Linear Conflict) with respect to node expansion count, execution time, peak memory, and solution depth. Features a Python-native Tkinter live demo and a high-performance C++ compiled solver backend. Directly applicable to robotics pathfinding and cybersecurity attack graph analysis.
