MLOps Pipeline• Architected and implemented an MLOps framework utilizing AWS Sagemaker for scalable model training, deployment, and monitoring, leveraging Docker for containerized workflows, and enhancing model and data governance practices to significantly optimize the machine learning lifecycle.• Integrated AWS services including Service Catalog, Step Functions, CodeCommit, CodeBuild, Lambda, Amazon ECR, and Bedrock with Guardrails, enforcing governance checks and compliance, and incorporated Whylabs as a model observability platform to create a robust and reliable pipeline.Deployed Applications to AWS• Successfully deployed applications to AWS ECS and Batch Fargate, leveraging Jenkins for continuous integration and continuous deployment (CI/CD)• Designed and implemented a PDF-conversion service to convert various document formats (images, Word) to PDF, integrating AWS Textract and Comprehend for efficient extraction for relevant information• Built a crawler service to extract relevant (as per the regex rule) news articles after crawling through multiple search engines like Bing search, Wikipedia, Glassdoor, Crunchbase, Zoominfo, etc.False Event Prediction: ML Model• Developed a ML algorithm to predict True or False events from a pool of crawled articles• Employed a range of machine learning classifiers such as AdaBoost, Bagging, Gradient Boosting, RandomForest, and SVM, finely tuned with GridSearch, culminating in a robust model with an impressive 98.7% accuracy rate.

RASA: Conversational AI(NLP)• Designed and implemented a comprehensive system including an API for querying SPLUNK logs, which identifies top movie and show searches using Postman, Flask, and MongoDB• Engineered a validation pipeline for enhancing a voice search RASA model, LLM, utilizing Docker, AWS Lambda, Step Functions, and EMR clusters, achieving a 3% increase in model accuracy.AWS Comprehend & AWS Transcribe: Sentiment Analysis• Executed real-time sentiment analysis of customer reviews in both written and video formats, leveraging Lambda functions, Snowflake, Kinesis, and Airflow for efficient data processing and analysis

Project ROCI• Engineered a graph database representing the logical topology of the access network, facilitating the creation of models and tools for proactive network health assessment and enhancement.• Architected an ETL pipeline using PySpark, Python, Scala, Pandas, EMR notebooks, Databricks, Delta Tables, GraphQL API, Lambda functions, and various AWS services, proficiently handling terabytes of data from multiple sources.• Utilized Gremlin query language for navigating the AWS Neptune Graph Database, enabling efficient traversal of access network nodes.Text Extraction-OCR• Built an OCR pipeline which utilizes Amazon Textract API (Python) for text extraction from unstructured scanned PDF documents• Saved around $800k/year by end-to-end automation of extracting specific field from AWS Textract JSON output leveraged S3, Lambda, Eventbridge, TF-IDF, CV, Word2Vec and Sagemaker• Led the team of RPA python developersNetwork System Monitoring• Developed a data pipeline for aggregating SNMP data from devices into Azure Data Lake, enabling analysis via Databricks and visualization with Grafana• Leveraged a suite of technologies including SNMP exporter, Telegraf, Prometheus, M3, InfluxDB, ADLS, Azure Event Hub, Kafka, and Grafana to facilitate efficient data collection, storage, and analysis

Led small teams of ML engineers and provided technical oversight and execution of various projects:Time Series Forecasting for Sales of a Group of Luxury Lifestyle Brands• Utilized FBProphet, and LSTM model for predicting weekly sales of in-store products using five years of sales data, using Naïve approach as a benchmark• Strategized a model for predicting sales of unreleased products using product similarityProtein-Disease Association for a Biopharma Startup developing Tissue Rejuvenation Therapeutics• Developed machine learning models to predict the role and association of proteins with age-related diseases (eg. Myopathy, Sacropenia)• Leveraged Machine Learning, Deep Learning, and QSAR methods (Pandas, Incremental PCA, t-SNE, KNN, Random forest and Autoencoder) to find the potential associationLoan Management for a Banking-Sector Client• Built various machine learning models to automate loan approval process, using data from past credit reports• Utilized AWS services (SFTP, EC2, S3, lambda, etc.) and machine learning methods (Pandas, SVM, Gradient Boosting and Random Forest) to automate the process

• Developed GROM software and webserver that helps predict mutations with 97% accuracy in tumor and regular genomes samples and works 70 times faster than any commercially available software• Leveraged Machine Learning skills and techniques (using Random Forest and Gradient Boosting, Flask, and HTML)• Automated regular testing and repetitive tasks using Selenium and Pyautogui

Built predictive machine learning models and algorithms for identifying ideal customer groups to be targeted for the marketing campaigns, and to find out the most significant factors that are truly involved in influencing the customer type (leveraged AWS, Lambda, S3)

• Programmed the infusion pumps to limit infusion during clinical use and minimize human errors• Utilized Ruby on Rails to write automation scripts for testing• Developed a search algorithm to extract information from log generated from infusion pumps field data using Python, Ruby on Rails, and Java• Provided computational support to research organizations• Developed an advanced pattern matching algorithm to identify certain sequences leveraging PERL and bash scripting• Corroborated in silico results with experimental and theoretical wetlab results