Projects / Systems / Proof

Building AI, robotics and backend systems with measurable engineering outcomes.

A technical portfolio focused on architecture, data flow, measurable performance and working implementation evidence.

Projects

5×

Faster

FPS

Computer Vision / AI System

1. Virtual AI Fitness Trainer

Final Year ProjectView GitHub

A dual-camera AI system for 3D deadlift analysis, combining pose estimation, stereo reconstruction, parallel video processing and rule-based biomechanics.

Problem

Single-camera form analysis loses depth and viewing-plane context, making deadlift faults such as back rounding and lateral tilt harder to measure reliably.

Technical Build

Engineered a dual-camera pipeline using OpenCV, MediaPipe, stereo reconstruction, rule-based biomechanics and a PyQt5 desktop interface.

Evaluation

Tested on 21 recorded sessions with 151 manually counted reps. The system detected 152 reps, giving 100.7% rep-count coverage.

Outcome

Achieved 85.7% of sessions within ±1 rep, 81.4% lateral-tilt precision, 92.7% lateral-tilt accuracy and 84.2% back-rounding precision.

Project Summary

Architecture, performance and evaluation

PythonOpenCVMediaPipePyQt5NumPyStereo VisionComputer Vision

Robotics / Navigation System

2. ROS2 Autonomous Robot Simulation

Simulation ProjectView GitHub

A ROS2 and Gazebo robotics simulation with autonomous navigation, object detection, odometry and landmark logging inside a maze environment.

Problem

Robots need to interpret their environment, react to traffic signs and navigate safely through a maze.

Technical Build

Built a ROS2/Gazebo simulation using YOLO detections, odometry, velocity control, obstacle-aware navigation and landmark logging.

Outcome

Connected perception, navigation and rule-based control into one explainable robotics workflow.

User Value

Shows how autonomous systems can make movement decisions from sensor, vision and map inputs.

Project Summary

Problem, build, outcome and user value

ROS2GazeboPythonYOLODockerNavigation

Cloud / Service-Oriented Backend

3. CycleNest Cloud Service

Backend SystemView GitHub

A Java-based cloud service using REST APIs, Apache Tomcat and MongoDB Atlas for location-based item retrieval and service orchestration.

Problem

Users need backend services that retrieve, process and return location-based data reliably through APIs.

Technical Build

Built REST endpoints using Java, Jersey JAX-RS, Apache Tomcat, Maven, MongoDB Atlas and Docker.

Outcome

Created a container-ready backend with structured API responses and cloud database integration.

User Value

Supports reliable communication between frontend requests, backend orchestration, external services and stored cloud data.

Project Summary

Problem, build, outcome and user value

JavaJerseyTomcatMongoDB AtlasDockerREST APIs

Big Data / Distributed Processing

4. Hadoop Traffic Data Analysis

Hadoop Project

A Hadoop-based distributed processing project that uses HDFS and MapReduce-style analysis to process large datasets and generate structured outputs.

Problem

Large datasets are difficult to process manually or on a single machine when the data grows in volume.

Technical Build

Used HDFS storage and MapReduce processing to split, process, shuffle and aggregate traffic data.

Outcome

Produced structured analytical outputs from raw data using a scalable big-data workflow.

User Value

Shows how raw operational data can be transformed into useful results through distributed computation.

Project Summary

Problem, build, outcome and user value

HadoopHDFSMapReduceJavaBig DataDistributed Systems

C++ / Scheduling System

5. Timetabling System

Software Engineering ProjectView GitHub

A C++ timetabling system that models users, sessions, rooms and scheduling logic to produce structured timetable records.

Problem

Manual timetable planning can become inconsistent when sessions, rooms and course constraints are handled separately.

Technical Build

Built a C++ application with separate models and modules for users, sessions, rooms and timetable operations.

Outcome

Created a structured scheduling workflow for creating, managing and viewing timetable information.

User Value

Helps organise timetable data clearly so users can manage academic scheduling with fewer manual errors.

Project Summary

Problem, build, outcome and user value

C++OOPSchedulingData StructuresCLI AppSoftware Design

Hybrid AI Assistant

6. GymAsis AI Chatbot

AI ApplicationView GitHub

A gym-focused AI assistant combining rule-based conversation, similarity search, voice interaction, cloud speech support, logic checks and image classification.

Problem

Fitness beginners often need fast answers on training, equipment and technique without searching through long resources.

Technical Build

Combined AIML rules, TF-IDF similarity matching, NLTK logic checks, Google Cloud voice support and a TensorFlow/Keras CNN.

Outcome

Created a multi-method assistant that handles direct questions, paraphrased queries, voice commands and image-based predictions.

User Value

The user can ask gym questions naturally, speak commands and classify gym equipment images inside one assistant workflow.

Project Summary

Problem, build, outcome and user value

PythonAIMLNLTKTensorFlowGoogle Cloud APITF-IDFCNN

Engineering Proof

Every project includes architecture diagrams, screenshots or working output evidence.

Measured Outcomes

Performance numbers are surfaced where available, including 5 min → 60 sec processing and 30 FPS capture.

Fast Navigation

Use filters, search or the project cards above to jump directly into the relevant system.

Back to Home