Vino Duraisamy

● Publishing articles on a variety of topics including Data Engineering, AI, ML, Statistics and Data Analytics.

lakeFS is an open-source tool that offers data versioning at scale for object stores - think "git for data", only it scales for 100s of petabytes datalakes. ● Start managing data the way you manage your code. ● lakeFS sits on top of your object store and provides git-like capabilities such as commit, branch, revert or merge all via UI, CLI or API. ● The ecosystem of tools you have today can access the data via lakeFS the same way they access it today. Easy peasy! Yep, I said it.

Tools & Languages Used: Python (Spacy), Named Entity Recognition Models (Decision Tree, Random Forest, CNN), Docker, Jenkins ● Developed and trained a Spacy based Named Entity Recognition model to identify and redact personally identifiable information from unstructured text data. ● Preprocessed and cleaned a corpus of 1M+ documents (emails, tweets, wikipedia and news articles) using Python. ● Engineered features for different document types and built an integrated feature pipeline for model training. ● Achieve an average F1-score of 80% for PII data and deployed the model as a python package in production. ● Worked with product managers and development teams to ensure accurate integration of the model into the product.

Tools & Languages Used: Python, Spark, Hive, AWS (S3, EMR, DynamoDB), Airflow, Hadoop, Docker, Jenkins ● Developed robust, scalable data pipelines for data privacy compliance module in Airflow for Customer Data Engineering & Analytics team at Nike. ● Created several DAGs to ensure hourly ingestion of customer data from AWS S3 buckets into Hive tables. ● Designed and built re-usable libraries in PySpark to ensure data quality and to support data engineering, and downstream analytics workflows. ● Continuously monitored and improved data pipelines by analyzing bottlenecks, and implemented efficient solutions. ● Refactoring & optimizing the code for improved reliability, performance, simplicity and maintenance.

Tools used: Python (SpaCy, Gensim, Regex), Prodigy, AWS S3 & EC2 ● Developed a prototype PII (Personally Identifiable Information) extraction and Named Entity Recognition model for Email corpus using Python. ● Leveraged SpaCy, Gensim and regex libraries for data cleaning and pre-processing to improve data quality. ● Built SpaCy based NER model and rule based PII extraction engine to identify, tag and redact PII data including health, finance, security and other sensitive information (such as Social Security Number, Bank Account Number, Credit Card number, etc. ) ● Re-trained SpaCy English language model ‘en_core_web_lg’ on Email corpus using the data annotation and training tool Prodigy for improved model performance.

● Worked with Dr. Asim Roy from Dept of Information Systems on improving explainability and interpretability of deep learning models. ● Built a 4 layer Convolutional Neural Network for MNIST handwritten digit recognition with 99.9% accuracy as the base model for analysis. ● Analyzed the filters, pooling layers and inter-connected layers of convolutional neural networks and visualized the activation values at each layer to understand underlying abstractions at each level. ● Leveraged Activation maximization to generate the input image that maximizes activations of particular neuron or a group of neurons, thus revealing the abstractions captured by the neurons. ● Used Saliency maps to understand the part of input image that contributes to the activations.

Tools used: Python, Tableau, MySQL, Excel ● Gathered data on customers, competitive products and market place, and consolidate information into actionable items, reports and presentations. ● Built a segmentation model to classify prospective customers using different factors enabling sales teams to derive efficient strategies for different prospect groups, thus decreasing the average sales cycle. ● Developed a customer churn prediction model to identify the top 3 factors contributing to customer churn, enabling the account managers to work on retention strategy. This increased the retention rate and improved the overall customer satisfaction score as well. ● Created Tableau dashboards and periodic reports for senior leadership to track customer churn metrics.