Created a cutting-edge machine learning model utilizing Retrieval-Augmented Generation (RAG), advanced LLMs, and vector storage to generate SQL queries, executing them, and providing results with human-interpretable summaries. This approach enhanced data retrieval and insight generation from complex databases seamlessly and enabled non-tech business analysts to interact with their systems.Deployed a Bayesian Marketing Mix Model to optimize ad spend and maximize ROI across channels including offline while analyzing saturation thresholds, cross-channel impact, and AdStock effects of marketing campaigns.

Built dynamic data pipelines for dynamic dashboards incorporating various data sources, including Google Places API as the backbone to compare retail outlets. Using fuzzy matching and sentiment analysis with NLTK and DistilBert for customer reviews to ensure accuracy of sentiments.

- Smart Campaigning using Multi-Class Classificationo Designed and developed a micro-segmentation strategy for customers of an international telco provider to run campaigns based on their historic usage of voice and data to augment up-sell and cross-sell of different packages. This included understanding of the existing purchase behavior and required building and deploying a multi-class classification model.o Analyzed and improved operationally the Data and Voice Churn Prediction models that were built by our team previously but were working rather slow. I worked with the HQL queries to improve joins and the overall speed of the script to make it more efficient.- Near Real-time Debit Card Fraud Detection & Preventiono Built and deployed the debit card fraud detection use case using Python as well as Watson Studio Modeler Flow to detect the integrity of transactions in near real-time for a leading bank in Pakistan. We improved the currently manual rule-based systems to complex decision trees. The model selected improved recall to 82% while reducing false positives to under 1000 a day.o A near real-time scoring mechanism was deployed as an API that was used to score transactions almost as soon as they were registered with the bank to see if they raise a flag and need investigation.

Worked closely with the clientele of IBM including global clients as a data science consultant to solve their business problems using data insights and KPIs using advanced analytics techniques applied on large and complex data sets of the clients.Using Machine-Learning and AI, we worked with IBM SPSS Modeler, IBM Watson Studio and Cognos Analytics suite to provide cutting-edge solutions to improve their business as well as customer experience while providing valuable insights on their key performance indicators and adding long-term business value. Alongside IBM tools, we leveraged Python as a part of the solution and potentially optimize the software application.I led the data science team including client interaction and steered end-to-end and timely use case implementations. I built and delivered the use cases independently while maintaining the internal team as well as client front. The use cases included Churn Prediction & Prevention, Next Best Offer, Customer Segmentation & Cash Optimization using algorithms including Time Series, Decision Trees, Neural Networks, SVM and Recommender System using Bayesian Algorithm and K-Means and Two-Step for Segmentation.I actively mentored data scientists inside and outside of IBM on approaching an industry-based problem, understanding the client and their need, tools for data science (Python & Watson Studio) and data science in general.Also worked as an Anti-Money Laundering Consultant for Leading Banks in Pakistan in both Functional as well as Technical Capacity. Maintained positive and growing relationships with the clients while delivering decided work on set timelines. Also enabled the Client by having UAT sessions and eventually pushing the final deployment to Production.

Assisted Students in delivering final projects for the course by helping them in debugging their issues or using technologies that can be useful to them in the course. Also worked with them to ensure exams and assignment preparation. This also included taking tutorials and relevant sessions.

Worked on multiple projects through the 2 month time span. Developed a web-based application for a Smarter Patient Management System with a team of highly talented and skilled developers using Node.js and relevant libraries and technologies.Developed a patch for the current application for a major client which enabled an upgrade for the client.Researched and worked on an image processing based counting algorithm to maintain and administer a transit system.

Assisted in Game Economics and Asset Balancing research for the Game in Development. Also helped and pushed in code whenever it was needed. Also includes experience in Unity3D Animations.