Data Science/Machine Learning Professional with 8+ years of experience in advanced analytics, data engineering, reporting and automation, proven expertise in leading data-driven initiatives, building machine learning and optimization models and delvering actionable insights to drive business performance. Strong track record in managing cross-functional teams and maintaining growth through advanced data science solutions.
- MLOps Tools: GCP Machine Learning Tools, BigQuery ML, TensorFlow, PyTorch
- Data Analysis & Visualization: Tableau, Power BI, Lifetime Value/ Customer Acquistion Cost Analysis
- Programming Language: Python, R, SQL, JavaScript, Advance Excel, Shell Script
- Database Systems & Cloud Computing: Amazon Web Services-Redshift, Athena, SQL Server, Google Cloud (Big Query), Azure Databricks
- Sector Focus: Digital Transformation, Technology, Media & Advertising, Investment Research
Co-Founder, PRL Insights & Partners (July 2024 - Present)
- Co-founded boutique data analytics and innovation based consulting startup that focuses on global macro based forecasting, quantitative finance & capital markets research using applied machine learning techniques.
- Implemented time-series techniques based on 90% sucess rate in back-testing & value at risk reduction (market risk) calculation for stock price volatlity.
- Constructed optimized portfolio to maximize performance while controlling risk, draw-down and trading cost, while enhancing the scalability & flexibility of portfolio
- Collaborated with data engineers to optimize feature engineering processes, improving model accuracy by +20% on average.
Manager - Data Science & Analytics, TD Canada Trust (_Nov 2021 - June 2024)
- Developed comprehensive measurement plans and performance scorecards to drive omni-channel strategies across owned channesl (owned & Paid Media), optimizing busines conversions, and implementing attribution models to achieve strategic business objectives.
- Established a daily automation process to transfer 500 GB 1TB of data from remote servers using seured socket shell (SSH) through putty terminals, storing them into in-house cloudera apache hadopp data lakes.
- Established e2e automated jobs for output data generation for email tableau reporting using Apache Ozzie job automation and scheduling operations saving 72 hours or resource time.
- Conducted a thorough cleaning of master campaign files to identify taxonomy gaps in reporting, utilizing an NLP algorithm. This process involved integrating data from various sources to ensure comprehensive and accurate reporting.
- Developed and implemented machine learning models to optimize digital marketing campaigns, leading to a +20% increase in ROI by enhancing ad targeting and audience segmentation.
Team Lead - Data Science Analytics, Epsilon (Jan 2019 - Dec 2020)
- Collaborated with cross-functional teams to design and implement data pipelines that automated the colelction, cleaning and analysis of digital marketing data, reducing reporting time by +30%
- Implemented machine learning algorithm based on +90% success on back-test using python code to predict customer behavioral insights, experience and loyalty.
- Implemented advanced machine learning algorithms and statistical techniques to analye digital media per formance, built measurement frameworks and attribution models to track performance across paid, earned and owned channels, increasing conversion rates by +15%
- Analyzed large sclae audience data (100M+ users) using Apache Hive and PySpark, enabling the optimization of media spend and personalized content recommendations for fortune 500 clients.
Manager - Data Scientist, iCrossing Company (Mar 2016 - Dec 2018)
- Developed and monitor quarterly sales forecast & optimization deliverable managing $29.4M investment
- Supervised 2 analysts on preparing & presenting monthly reports, implemented python code & APIs integration by reducing reporting cycle by 20%
- Constructed a seasonal forecast model managing $16M investment resulting in +15% ROI improvement
- Performed click-stream data analysis on data stored in Amazon AWS S3 to uncover consumer behavior patterns and key variables influencing conversion, delivering real-time insights into engagement metrics.
- Performed sentiment analysis on customer reviews and social media data, identifying key trends that informed content creation and social media strategy, driving a +25% increase in engagement.
Senior Data Analyst - Nielsen (Mar 2015 - Mar 2016)
- Collaborated with cross-functional teams to gather requirements and created detailed data warehouse schematics, field specifications and data management processes to improve operational efficiency by +20%
- Optimized client overall marketing budget by saving 4.5% $ margin on digital and traditional channels using media saturation curves built through marketing planners.
- Developed machine learning models to optimize tag configurations, enhancing conversion rates by analyzing tag-triggred events and customer segments
- Implemented machine learning pipelines and automation tools, reducing manual reporting time by 30%
The concepot of RFM (Recency, Frequency and Monetary) in market research extent way back when historical transactions started to be collected in databases. Suddenly, we had DB-managers and analst looking into databases to understand response rate performance to evaluate direct marketing campaigns effectiveness. By simply looking into co-horts based on RFM, analysts were able to notic stark changes in customer responses to similar campaigns. So, in order to improve performance, they would mail customers more frequently who are more likely to respond than others.
MMM is a statistical analysis to estimate the past impact and predict the future impact of various marketing tactics on sales, can deeply inform marketing plans. I created a full stack econometric model for marketing mix to evaluate the sales forecast, it is important to understand the concept of advertising stock (adstock) decomposition for campaigns, the level of reach, penetration and frequency ordering to manage various media channels. We would also go into details of how sales distribution and price-promotion plays role in creating a full stack economic models.
Marketers in all media are interested in the concept of optimal frequency, few know how to actually measure it. Coommon sense tells us that too few impressions won't generate significant impact, while too many impressions delivered to a given user results in over-saturation and a waster of media dollars.
- Causality: The new science of an old question - GSP Seminar, Fall 2021
- Guest Lecture: Dimensionality Reduction - Big Data and Machine Learning for Scientific Discovery (PHYS 5336), Spring 2021
- Masters, Engineering Industrial & Statistics | The Texas A&M University
- Bachelors, Engineering Mechanical | Visvesvaraya Technological University