Machine Learning Computer Vision

Manav Desai

email : *****.*****@*******.*** LinkedIn : https://linkedin.com/in/manavdesai22

Phone : 862-***-**** Location : Philadelphia, PA Machine Learning and Data Science engineer with 2.5 years of experience building end-to-end machine learning pipelines, predictive models, and real-time computer vision systems. Proficient in Python, SQL, and deep learning; experienced in deploying models, conducting exploratory data analysis, and driving business strategies and impact through data-driven decision-making SKILLS

Languages: Python, R, SQL, MATLAB, C++

Libraries & Frameworks: scikit-learn, PyTorch, TensorFlow, XGBoost, YOLOv8, spaCy, OpenCV, Pandas, NumPy, Matplotlib, Seaborn, Keras Tools & Platforms: Tableau, Power BI, Streamlit, Docker, Git, Jupyter, Roboflow, Hive, Hadoop, Excel, AWS, OpenCV, SaaS Soft Skills: Clear empathic communicator, efficient time manager, collaborative team leader and player, resilient under pressure EDUCATION

Drexel University Philadelphia, USA Sep 2022- Jun 2024 Master of Science in Robotics and Automation (Artificial Intelligence / Machine Learning), GPA: 3.4/4 Manipal University Jaipur Jaipur, India Jun 2018 - Jun 2022 Bachelor of Technology in Mechatronics

EXPERIENCE

Machine Learning Research Assistant : Drexel University Philadelphia, PA, USA Oct 2024 - present

●Implemented a dual-drone computer vision system with YOLOv8 for soccer tracking instead of GPS, reducing player tracking costs by 40%

●Applied heatmap analysis and time-series analysis to optimize in-game strategies, improving efficiency by 15% and in-game adjustments by 20%

●Built multivariate regression models on 100+ player metrics to forecast match outcomes with 25% accuracy gain

●Improved team performance metrics by 25% through detailed predictive analytics of player movements and formation efficiencies Computer Vision Engineer Co-op : LightBoxVR San Francisco, CA, USA Jun 2023 - Jan 2024

●Created real-time detection models improving accuracy by 18%; enhanced preprocessing to boost image model performance by 20%

●Automated model retraining via CI/CD pipeline using Docker and Git, cutting error rate by 25%

●Enhanced image preprocessing for better feature extraction from rotated and skewed images, boosting performance by 20% ML and Data Analyst : Larsen and Toubro Hazira, GJ, INDIA Jan 2022 - May 2022

●Developed forecasting models using Python and SQL to predict abnormal energy consumption patterns, leading to anomaly detection and reducing costs by 10%

●Implemented CNN-based thermal anomaly detector that reduced energy costs by 15%

●Partnered with control systems engineers to integrate AI insights into SCADA dashboards for real-time operational visibility

●Delivered analytical dashboards using Excel and Power BI to uncover KPI, which reduced downtime by 15% Junior Data Scientist : Aniket Metals Mumbai City, MH, INDIA Jan 2021 - Dec 2021

●Designed supervised models (SVM, Logistic Regression) to flag product defects with 86% accuracy

●Conducted feature engineering and classification analysis on manufacturing data to identify high-risk defect patterns, enabling proactive quality checks and reducing post-production rework by 22%

●Integrated MES pipeline to enable real-time QC alerts, increasing yield by 18% and decreasing batch variability by 31% ACADEMIC PROJECTS

Vehicle Detection & Counting for Smart Traffic Systems with Computer Vision Drexel University Mar 2024 Techniques: YOLOv8, Object Tracking (Deep SORT), Python, OpenCV, Heatmap Generation, Time Series Analysis

●Built a vehicle detection and tracking system using YOLOv8 and Deep SORT, achieving 90% tracking accuracy on intersection footage

●Analyzed congestion trends via heatmap analysis and time series forecasting, which can help city planners on signal optimization patterns Analyzing Philadelphia Crime Data Drexel University May 2023 Techniques: Data Cleaning, K-Means Clustering, Time Series Analysis, Geospatial Mapping, Tableau, Python, Excel

●Processed 618K+ crime records using Python (Pandas, NumPy) and Excel, ensuring data integrity through imputation and outlier removal

●Applied K-Means clustering and geospatial mapping (Folium, Seaborn) to identify high-risk areas; used time-series decomposition and box plot analysis to reveal temporal crime patterns

●Developed interactive Tableau dashboards for spatial and temporal drill-downs, empowering stakeholders with actionable public safety insights Credit Card Fraud Detection System Drexel University Mar 2023 Techniques: Logistic Regression, XGBoost, SMOTE, Confusion Matrix, ROC-AUC

●Designed a binary classification pipeline using Logistic Regression and XGBoost on imbalanced credit card transaction data; achieved 92%

●Applied SMOTE oversampling and feature selection to improve model precision and reduce false positives by 18%, enabling real-time flagging of suspicious activity

Sales Forecasting and Inventory Optimization Manipal University Jun 2020 Techniques : Linear Regression, Time Series Forecasting, Power BI, Trend Analysis, Demand Planning, Python (pandas, statsmodels)

●As part of an academic capstone, collaborated on a retail data project to forecast monthly sales using linear regression and time-series modeling, leading to a simulated 22% reduction in overstock

●Created Power BI dashboards to visualize demand trends across regions and provided inventory planning recommendations based on forecast accuracy

Contact this candidate