Sania Jawaid

Partnered with cross-functional leaders to define and execute AI/ML strategy, aligning technical initiatives with business growth and customer value. Led end-to-end development of advanced ML and AI solutions, from ideation to deployment, ensuring scalability, reliability, and measurable business impact. Mentored teams (8-9 people) on modern data science practices, fostering adoption of AI-driven tools and state-of-the-art research. Developed and deployed a CNN-based EV detection system (F1 Score 0.96), scaling to 100K+ residential consumers via 1TB automated data pipelines (PySpark, Databricks, Airflow), driving $1.76M in new revenue. Built a RAG-powered chatbot that enabled real-time facility issue resolution, reducing critical outage costs ($5.6K per minute). Increased anomaly detection accuracy to 96% using an LSTM-based Quantile Regression Model, cutting false alarms to less than 3 per month across 400+ IoT devices. Delivered a fault prediction ensemble model (TCN, GRU, XGBoost) with 94% accuracy and 30-day advance warnings. Designed an Agentic AI framework leveraging MCP server and Azure AI services to automate the SDLC, supporting legacy code migration (Java, C++, Pascal, Fortran) and new feature generation with 45% effort reduction. Reduced testing time by 40% by creating an LLM-powered Pytest scaffolding framework for auto-generating test cases from source code.

Designed and deployed a deep learning encoder-decoder system to predict telecommunication bandwidth shifts by conversion of HLog from 17MHz to 106MHz and 121MHz, enabling fault detection across 40M+ signals and improving network reliability and performance.

Built a computer vision-based grain loss detection tool, achieving 85% accuracy at 0.5% detection threshold, directly helping farmers minimize harvest wastage and optimize yield.

Optimized large-scale similarity search via product quantization for efficient KNN. Applied multi-class deep learning for semantic segmentation, PCA-based face recognition, and spam filtering with online learning. Executed Big Data projects with Spark/Hadoop, Map-Reduce, for trip reconstruction and itemset mining. Solved optimization problems (e.g., Travelling Salesman) using Genetic Algorithms.

Led 47 member teams to deliver global client applications like CarDekho Auctions, serving 1,000+ users with scalable design, realistic planning, and effective risk management. Led teams to deliver enterprise applications with AI-driven personalization and analytics. Architected scalable full-stack platforms (frontend, backend, DevOps) while driving technology adoption and mentoring team members. Built a hybrid recommendation engine (content-based + collaborative filtering) to suggest relevant patient matches, forums, and articles using TF-IDF/Word2Vec embeddings and user-content interaction signals. Used TF-IDF and classifiers (SVM/Logistic Regression) to tag automatically and route student queries to the right subject/domain expert.