Tensorflow
ANANDA RAO MIKKILI
Lead Data Scientist
Houston, TX,KA
Summary
Certified Lead Data Scientist with profound technical skills having over 12 years of experience is looking for opportunity to work on MLOps or entire Machine/Deep Learning lifecycle to deliver business value through scheduling, building, deploying and monitoring accurate machine learning models
Overview
13
13
years of professional experience
6
6
years of post-secondary education
2
2
Certificates
4
4
Languages
4
4
Years of experience in Automating AI/ML workflows (CI-CT-CD))
4
4
Years of leadership experience for AI/ML team
8
8
Years Experience in Machine Learning
Work History
Data Scientist Manager
6d Technologies (Client: Tigo - Guatemala)
05.2020 - 01.2023
- Technologies & Cloud Services : AWS MWAA (Managed Workflow Apache Airflow), AWS SageMaker, AWS Glue, AWS Kinesis, AWS CodePipeline, AWS Lambda, AWS API Gateway, AWS ECS/Fargate, Tensorflow, Scikit-Learn, Tableau, PySpark, Python, AWS S3, AWS Redshift, AWS DynamoDb, AWS CloudWatch, Git
- Building and maintaining CI-CT-CD pipelines for Deep Learning based workflows [Product Recommendation Engine and Churn Propensity Prediction & Management] on AWS MWAA (Managed Workflow Apache Airflow) or AWS SageMaker using Python and PySpark
- Achieved an accuracy of 72% of for product recommendation and 85% for churn propensity prediction in production
- Enhanced revenues for client upto 7% overall through recommendation engine deployed and able to retain 59% of predicted churners through freebie campaigns
- Delivered highly accurate models in the production for Product Recommendation Model & Churn Propensity Prediction Model
- Perfomed Data Analytics to derive and understand customer behavioural patterns and identify key data segments as part of feature engineering
- The models have been monitored using AWS SageMaker and AWS CloudWatch to validate model performance
- Performed A/B Testing for smooth movement of productionizing the models from SIT environment after appropriate validation of new model when compared with existing production model
- Model training, tuning and building has been performed using Tensorflow through the process called Build Your Own Model (BYOM) on AWS environment
- Data pipeline has been created using AWS Glue from AWS Redshift. The data transformations have been created for deriving new segments out of existing segments from AWS Redshift database
- Model Metrics for every iteration (epoch) has been captured into AWS DynamoDB for visualization over Tableau/OmniBoard. The experimentation results are also stored in DynamoDB to compare and showcase to client whenever required
Senior Data Scientist
6D Technologies (Client: OmanTel - Oman)
01.2018 - 03.2020
- Technologies & Cloud Services : AWS MWAA (Managed Workflow Apache Airflow), AWS SageMaker, AWS Glue, AWS Kinesis, AWS CodePipeline, AWS Lambda, AWS API Gateway, AWS ECS/Fargate, Tensorflow, Scikit-Learn, Tableau, PySpark, Python, AWS S3, AWS Redshift, AWS DynamoDb, AWS CloudWatch, Git
- Building and maintaining CI-CT-CD pipelines for Deep Learning based workflows [Product Recommendation Engine] on AWS MWAA (Managed Workflow Apache Airflow) or AWS SageMaker using Python and PySpark
- Delivered highly accurate models in the production for Product Recommendation Model
- Perfomed Data Analytics to derive and understand customer behavioural patterns and identify key data segments as part of feature engineering
- The models have been monitored using AWS SageMaker and AWS CloudWatch to validate model performance
- Performed A/B Testing for smooth movement of productionizing the models from SIT environment after appropriate validation of new model when compared with existing production model
- Model training, tuning and building has been performed using Tensorflow through the process called Build Your Own Model (BYOM) on AWS environment
- Data pipeline has been created using AWS Glue from AWS Redshift. The data transformations have been created for deriving new segments out of existing segments from AWS Redshift database
- Model Metrics for every iteration (epoch) has been captured into AWS DynamoDB for visualization over Tableau/OmniBoard. The experimentation results are also stored in DynamoDB to compare and showcase to client whenever required
- Achieved an accuracy of 65% of for product recommendation in production and able to make subscribers impressions on the recommended products
- Able to lift the revenue of 4% - 6% overall through recommendation engine deployed
- Collaborated with client for requirement gathering, sprint plan, sprint review, sprint retrospection
- The time complexity of the entire pipeline execution has been reduced from 16 hours to 5 hours through code optimization and applying parallel processing (multi-threading) wherever possible
Senior Data Scientist
6D Technologies (Client: BTC - Botswana)
06.2017 - 04.2018
- Technologies: Apache Airflow, Seahorse (deepsense.ai), Tensorflow, Keras, Pandas, Numpy, Scikit-Learn, R, Python, Zeplin/Streamlit, Plotly, GitLab, Git, Pytest
- Building CI-CT-CD pipelines for Churn Propensity Prediction & Management, Product Recommendation Engine, Statistical Subscriber Segmentation using RFM and Revenue Forecasting on Apache Airflow and Seahorse Deepsense using python (In house server deployment)
- Performed Data Analytics to derive and understand customer behavioral patterns and identify key data segments.
- Data pipeline has been created using Seahorse (deepsense.ai) from Hive & MySQL databases. The data transformations have been created for deriving new segments out of existing segments and created training as well as test datasets
- Delivered highly accurate models in the production for Churn Propensity Prediction Model, Product Recommendation Model, Statistical Subscriber Segmentation using RFM and Revenue Forecasting Model
- Performed Stratified Sampling to divide entire subscriber base into Control Group and Target Group where A/B Testing can be performed for campaigns launched by comparing revenue lift from both Control Group and Target Group
- Collaborated with client for requirement gathering, sprint plan, sprint review, sprint retrospection apart from daily scrum with team
- Models have been monitored manually (sporadic base test data) and measured the degradation of models
- Knowledge Transfer sessions have been delivered (by me) on code walk through, implementation procedure, scheduling and monitoring. These models can be scaled up as subscriber base increases
- Predicted results have been loaded to database, analyzed and visualized to measure accuracy by comparing with actual data after a period of time
- Created all the documents from the beginning like HLD, LLD, Architecture and shared with client team
Data Scientist
6d Technologies (Client: Cell C - South Africa)
01.2016 - 07.2017
- Technologies: Seahorse (Deepsense.ai), Tensorflow, Keras, Pandas, Numpy, Scikit-Learn, R, Python, Zeplin/Streamlit, Plotly, GitLab, Git
- Building CI-CT-CD pipelines for Churn Propensity Prediction & Management, Product Recommendation Engine on Seahorse (deepsense.ai) using python and PySpark (server deployment at client location)
- Data pipeline has been created using Seahorse (deepsense.ai) from Hive & MySQL databases. The data transformations have been created for deriving new segments out of existing segments and created training as well as test datasets
- Performed Stratified Sampling to divide entire subscriber base into Control Group and Target Group where A/B Testing can be performed for campaigns launched by comparing revenue lift from both Control Group and Target Group
- Delivered highly accurate models in the production for Churn Propensity Prediction Model and Product Recommendation Model
- Performed Data Analytics to derive and understand customer behavioral patterns and identify key data segments
- Models have been monitored manually (sporadic base test data) and measured the degradation of models
- Predicted results have been loaded to database, analyzed and visualized to measure accuracy by comparing with actual data after a period of time
- Created all the documents from the beginning like HLD, LLD, Architecture and shared with client team
Senior Software Engineer
Altisource Business Solutions
08.2013 - 12.2015
- Technologies: Java/J2EE, JUnit, Apache Tomcat, GitLab, Jenkins, Spring, RESTful Web services, Hive, HBase, MySQL, MongoDb, Maven and SonarQube
- Designing and development of enterprise mortgage platform (Loan Resolution Module) for HAMP and Non-HAMP models for JP Morgan & Chase to help underwriters to process the mortgage loans. The project is completely built on Spring Boot and its services and deployed on hosting servers at premise
- Designing and development of Servicing Management Default Underwriter (SMDU) for mitigating loss against property loan through building Automated Valuation Models
- Deployment of these platforms over application servers and measured performance with respect to database queries, API calls
- Performance improvement has been done for LRM and SMDU through applying Multi-Threading. Performance has been improved to 5 times
- Taking care of build release management to automate entire CI-CD process over GitLab - Jenkins - App Servers
- Involved in bug fixing and management, to interact with quality assurance team to cross-verification and resolving bugs raised against platforms
- Written unit test cases against functionalities developed for LRM and SMDU
- Performed code quality and static code analysis for coverage, defects and memory leaks by using SonarQube
Software Engineer
Philips Electronics
Bangalore, Karnataka
05.2011 - 08.2013
- Technologies: Java/J2EE, JUnit, Apache Tomcat, SVN, Jenkins, Spring, RESTful Web services, Maven and CodeCover
- Worked with software development and testing team members to design and develop high performance, scalable and robust health care platform called Shushrutha (Cardio Vascular Procedural Machine)
- Developing automated code coverage framework to execute and measure code coverage and memory leaks for Shushrutha
- Involved in fixing bugs and build release management for Shushrutha
- Written LLD for some modules in Shushrutha
- Written unit test cases for entire Shushrutha, executed test cases and generated unit-test case reports
- Interacted with quality assurance team to cross-verification and resolving bugs raised against Shushrutha
Technology Engineer
Tesco Hindustan Service Centre (Tesco PLC)
Bangalore, Karnataka
07.2010 - 05.2011
- Technologies: Java, JUnit, Apache Tomcat, SVN, Android and Maven
- Part of development team for e-commerce platform for Tesco PLC in both web based application as well mobile application (Android based)
- Part of development team for Ingestor Engine, system which is used to ingest and process XML files using multi-threading and load results to database (Mongo DB)
- Written unit test cases using JUnit for functionalities developed for both Android mobile app and Ingestor Engine
- Implemented Multi-Threading (parallelism strategy) for Ingestor Engine to reduce time taken for transformation and data loading to MongoDB
Education
Master of Science - Telecom Technology And Management
Indian Institute of Technology, Delhi
New Delhi 07.2008 - 08.2010
Bachelor of Science - Electronics & Communications Engineering
Andhra University
Visakhapatnam 08.2002 - 04.2006
Skills
Machine Learning Operations (MLOPs)
Applied Machine Learning
Deep Learning
Data Segmentation/Clustering
Data Analytics - Identifying Behavioral Patterns
Advanced Data Visualization
Model Monitoring
Performing A/B Testing for ML Models
Performing Feature Engineering
Expertise in Python
Expertise in PySpark
Expertise in R
Automating ML Workflows (CI-CT-CD)
Building Recommendation Engines
Building Propensity Prediction Models
Building Regression based Estimation Models
Automated Data Pipelines for Machine Learning
Building ETL Pipelines for most of the databases
Accomplishments
- Achieved significant revenue uplift by designing, developing and delivering ML based CI-CT-CD Pipeline using AWS MWAA for product recommendation system entirely on AWS. The scheduling workflows, model monitoring and performance visualization capabilities have been enabled for TIGO Guatemala with highly accuracy models
- Successfully led and supervised the team of 10 in the development of AI/ML driven MAGIK (CVM platform) for a period of 5 years
- Collaborated with cross functional teams across the globe
- Created and achieved signed-off all the documents (SSD/HLD/LLD..etc.) with clients through proficient documentation and clear content
Software
PyTorch
Keras
H2Oai
Pandas
Numpy
Matplotlib
Seaborn
AWS MWAA (Managed Workflow Apache Airflow)
AWS SageMaker
AWS CodePipeline
AWS ECS/Fargate
AWS Glue
AWS EMR
AWS Kinesis
AWS Lambda
AWS API Gateway
AWS CloudWatch
AWS Redshift
AWS DynamoDb
AWS S3
Streamlit
GitLab/Git
Seahorse Deepsense
Tableau
MongoDB
Hive
HBase
Oracle
MySQL
Certification
Microsoft Technology Associate: Introduction to Programming using Python
Interests
Exploring Computer Vision AI solutions over GPU [RAPIDS]
MLOPs over Cloud Environments like AWS, GCP & Azure
Expertise over Advanced Function Programming Concepts
Explainable Machine Learning
Effortless Data Visualization
Seemless Integration of Data Driven System
Affiliations
- Member of Medium (https://medium.com/@anandmikkili)
- MLOps (https://www.linkedin.com/company/mlops-world/)
Timeline
Microsoft Technology Associate: Introduction to Programming using Python
07-2021
Data Scientist Manager
6d Technologies (Client: Tigo - Guatemala)
05.2020 - 01.2023
Advanced Machine Learning with Python
03-2020
Senior Data Scientist
6D Technologies (Client: OmanTel - Oman)
01.2018 - 03.2020
Senior Data Scientist
6D Technologies (Client: BTC - Botswana)
06.2017 - 04.2018
Data Scientist
6d Technologies (Client: Cell C - South Africa)
01.2016 - 07.2017
Senior Software Engineer
Altisource Business Solutions
08.2013 - 12.2015
Software Engineer
Philips Electronics
05.2011 - 08.2013
Technology Engineer
Tesco Hindustan Service Centre (Tesco PLC)
07.2010 - 05.2011
Master of Science - Telecom Technology And Management
Indian Institute of Technology, Delhi
07.2008 - 08.2010
Bachelor of Science - Electronics & Communications Engineering
Andhra University
08.2002 - 04.2006