QuinStreet

• Ownership & Business Impact: Technical owner of the end to end ML ranking architecture for the Personal Loans vertical; scope expanded in 2026 to include the conversion model alongside CTR models, owning ranking performance, deployment strategy, and experimentation.
• Technical owner of Personal Loans ML ranking for 3M+ annual inquiries, driving expansion from subprime to higher quality traffic, about a $2M margin opportunity.
• Built the Personal Loans XGBoost vendor conversion model and own the Personal Loans CTR model, with weekly cron retraining; a persistent 90/10 holdout against the legacy lookup baseline shows about a 10% profit lift on a $1M+ margin segment.
• Lead a margin preservation initiative combining CTR and CVR predictions to filter unprofitable leads before purchase, projecting an 8 to 12% reduction in wasted lead spend.
• Own the Cython simulation framework, which stress tests ranking strategies before release and monitors Auto Insurance production health daily; cut runtime by 50% and maintain parity with live production logic.
• Help onboard new algorithm engineers on ranking architecture and code reviews; authored team code quality and model release standards, now followed team wide.
• Led cross functional design reviews for the next generation ranking engine, aligning data science objectives with infrastructure constraints.
• Tools & Techniques: Python, XGBoost, Scikit-learn, Cython, Redshift, MySQL, Feature Engineering, SHAP, A/B Testing, Cross validation, GBM, Ranking Models

QuinStreet

• Built and productionized a conversion prediction model for the Personal Loans funnel, complementing CTR models to optimize downstream conversion.
• Optimized an automated simulation testing suite using Cython, achieving 95% fidelity with production environments and significantly improving test coverage and computational efficiency.
• Engineered high impact feature sets using domain specific aggregations, temporal windows, statistical summaries, and one hot encoding—improving model generalization and reducing overfitting.
• Implemented advanced supervised learning models (e.g., Gradient Boosting Machines, Logistic Regression), using Bayesian optimization and cross validation to fine tune hyperparameters and boost ranking accuracy.
• Conducted in depth feature importance analysis (e.g., SHAP, permutation importance), removing low signal features without sacrificing performance.
• Delivered production ready machine learning pipelines in collaboration with product and data engineering teams, enhancing metric tracking for advertiser engagement.
• Implemented the Hungarian algorithm for optimal constrained ranking assignment, validated against the greedy baseline through staged online A/B tests, lifting margin revenue 2.5 to 8% depending on traffic; now live in Auto Insurance and Banking.
• Tools & Techniques: Python, Scikit-learn, Cython, Redshift, MySQL, Feature Engineering, SHAP, A/B Testing, Cross validation, GBM, Hungarian algorithm, Data Pipelines, AdTech, Ranking Models

VidMob

• Built deep learning models in Python to score creative ad assets and support automated engagement prediction and recommendation pipelines within an advertising analytics platform.
• Designed and implemented deep learning models for computer vision tasks—such as object detection and image scoring—resulting in a 25% increase in the accuracy of ad engagement classification.
• Applied supervised learning and statistical modeling techniques to predict user interaction rates, followed by rigorous A/B testing and performance benchmarking.
• Engineered modular, reusable components for data preprocessing, model training, and evaluation, reducing pipeline execution time and improving development efficiency across ML iterations.
• Collaborated with cross functional teams (data, product, and engineering) to deploy scalable AI driven insights into production through predictive analytics.
• Tools & Techniques: Python, PyTorch, Scikit-learn, Computer Vision, Deep Learning, MySQL, Git, SQL, A/B Testing, Feature Engineering

Accenture

• Automated pharma sales ETL workflows in Python, SQL, and Apache Hive, cutting analysis time by 50%.
• Built drug sales analytics and Tableau dashboards, lifting competitor match rates by 20%.
• Onboarded 10 graduate engineers.
• Global Innovation Challenge Semifinalist at Accenture.
• Tools: Python, SQL, Apache Hive, Tableau

VMG

• Wrote 2 chapters for a book on Computational Imaging, which was published by MIT Press.
• SIGGRAPH 2022 spotlight: Blending Camera and 77 GHz Radar Sensing for Equitable, Robust Plethysmography.
• Collaborated with a team of engineers to develop and implement a state of the art Shift Robust Loss Function for rPPG, resulting in decreased error by 40%
• Worked with team to diagnose skin tone bias in medical application using multimodal fusion between radar and RGB data. As a result, the team was able to develop an algorithm that improved accuracy by 75%. This work is the SIGGRAPH 2022 publication listed above.
• Tools: Python, OpenCV, Keras, Tensorflow, PyTorch, Computer Vision, Computational Robotics, Scipy

Indian Institute of Science (IISc)

• Collaborated with Dr. Sundaram and a team of researchers to pioneer new methods for faster and more accurate diagnosis of Autism.
• Successfully cleaned and rearranged ABIDE Dataset (depending on correlation between different regions of the brain) to improve data correlation, leading to more accurate predictions.
• Developed and optimized deep neural networks to analyze fMRI scans for early detection of Autism, achieving an 80% accuracy rate on a large set of cleaned data.
• Code: GitHub repo. Report: Project report.
• Tools: Python, PyTorch, TensorFlow, OpenCV, MATLAB, Research, Scipy, Pandas

UCLA

• Teaching Associate at UCLA: led 2 discussion sections and 2 lab sessions, guiding students in mathematical modeling in Python.
• Evaluated student work, gave feedback on assignments, and supported faculty with course preparation and grading.

Ranking and Experimentation

Languages and Core Tools

ML and Scientific Libraries

Data and Cloud Tools