- Center of Intelligent Systems and Networks Research (June 2017 – Present)
- Center of Intelligent Systems and Networks Research (Jan 2016 – May 2017)
Undergraduate Researcher and Developer
- University of Engineering and Technology, Peshawar (Sept 2013 – May 2017)
Electrical Engineering (CGPA 3.52/4)
Machine Learning/Artificial Intelligence
- Data Driven Microgrid: (Machine Learning, Data Analytics, Energy Informatics, Python/Matlab)
Load consumption Dataset acquired by OpenEI.org, Solar/Wind data acquired from NSRD, wunderground.com. Performed Time-Series Predictive Analysis of Solar/Wind Power Production and Power Consumption using LM-Q algorithms, Bayesian optimization and CGPANN. Further used the statistics to form a Data Driven Microgrid that took real-time decisions on Energy consumption/storage to mitigate consumer expenses.
- Spam Classifier: (Machine Learning – NLP, Python)
Dataset acquired from SpamAssassin Public Corpus i.e spam_2, easy_ham_2, hard_ham training set. Built vocabulary list of emails after Stemming, Tokenization, stripping html tags and normalization. Trained/cross-validated/tested Support Vector Machines cross-validated resulting in a spam classifier of 96.5% accuracy.
- MINST Handwritten Digit Recognition: (Machine Learning – Computer Vision, Matlab)
Used back-propagation to train an artificial neural network over a training set of 5000 images of handwritten digits, achieving 99.6% accuracy for a test set of images. Vectorized (optimized) implementation of the Neural Network.
- Cartesian Genetic Programming to Design Boolean Functions: (AI – Bioinformatics)
Used Cartesian genetic programming to develop a program that can design a digital circuitry for an infinite amount of inputs and outputs as required by the user.
- Kaggle Titanic – Machine Learning From Disaster: (Machine Learning – Matlab)
Created a new feature for passengers with same family name. Assigned mean fares to missing values according to cabin class. Missing embark station assigned as most frequent one. Implemented Random Forest technique for prediction model – accuracy 84.7%.
Sensor Networks/Embedded Systems
- Smart Metering: (C, PIC32, PHP, MySQL GSM module)
Developing a 12 bit resolution smart meter using Shunts/Op-amps for live readings from a High Tension Power Transformer. Debugging and improvement of smart meter architecture. Data Analysis and Visualization of Smart Meters.
Simcom GPS module interfaced with microcontroller which sent data to the webserver via a GSM module. The webpage had google maps api embedded that showed a live feed of the location of the tracker.
- Online Courses: Machine Learning by Stanford University on Coursera – Certificate earned on August 1, 2016