Worky | Meet our Talent - Sai krishna Bachu

Skills

Machine Learning (ML)

About

BACHU SAI KRISHNA's skills align with System Developers and Analysts (Information and Communication Technology). BACHU also has skills associated with Programmers (Information and Communication Technology). BACHU SAI KRISHNA has 9 years of work experience.

Work Experience

Machine Learning Engineer IV

IGT PLC

July 2021 - Present

AI Model Deployment & Monitoring on Azure AKS using Nvidia Triton: Initiated the implementation of Triton Inference Server with Python on Azure Kubernetes Service (AKS), overseeing the efficient deployment of Summarizer and Labeler models developed by data scientists. Leveraged Triton's advanced capabilities for model serving, including dynamic batching and model versioning, to optimise inference performance and scalability. Developed robust APIs for interacting with the deployed models, ensuring seamless integration into production environments while maintaining high reliability and performance standards. Additionally, implemented comprehensive monitoring solutions to track system performance, model inference metrics, and resource utilisation. AI Services 2.0: Designed and implemented a microservice architecture for deploying machine learning models seamlessly from training to production on Azure Kubernetes Service (AKS). Developed domain-driven microservices to serve multiple internal teams efficiently. Analysed data requirements for these services and established a configuration layer for tables. Engineered precomputed jobs to mitigate on-the-fly computations, thus enhancing service performance. Orchestrated all services using Docker and implemented GitHub Actions for streamlined container registry pushes. Established a CI/CD pipeline to automate the build and deployment of Docker images into the AKS cluster. Automated ML Pipeline: Designed and implemented a batch ML pipeline automating feature engineering, training, inference, and monitoring processes for diverse ML models. Successfully productionized "Recommendation Engine" and "BTYD" models. Established a robust data environment for ML and predictive analytics. Leveraged AWS services including EventBridge, Lambda, SageMaker, SNS, S3, Glue, Athena, API Gateway, and QuickSight to construct the entire architecture. Analysed ML model outcomes and crafted Quicksight dashboards for marketing and product teams. Engineered APIs to deliver ML model outcomes to data scientists and web applications. Configured and deployed the entire architecture using Terraform scripts for seamless automation. Data Visualisation Object Migration Pipeline: Designed and implemented the BIOps architecture to facilitate the migration of visualisation-related objects and enable version control for the analyst team. Engineered an on-demand migration solution to transfer specific assets across AWS accounts efficiently. Implemented a version control mechanism to monitor changes in dashboards developed by analysts. Developed essential functions such as DescribeObject, CreateObject, UpdateObject, DeleteObject, and ListObject to streamline the migration process. Constructed the entire architecture utilising AWS services including QuickSight, Lambda, and API Gateway. Configured and deployed the complete architecture using Terraform scripts for seamless automation and management.

Data Engineer (CoE)

SS&C EZE Intralinks

February 2020 - July 2021

AI NER Redaction Pipeline: Implemented an AI NER pipeline to redact the sensitive PII information in different documents stored in legacy applications. Developed a text extraction service which extracts the text and is stored in a data store. Provided AI preprocessing/vectorization service subscribes to text extraction complete events and performs data preprocessing activities such as data wrangling, cleanup, tokenization etc. on the raw text and saves it into vectorized data lake (VDL). Designed and implemented an orchestrator service which is responsible for batch creation, triggering pipelines. Modelled the database to store the ML model results to support real time users. Data Streaming Platform: Designed and developed producer, consumer services for storing, retrieving multiple events. Implemented the data streaming component to transform the data on the fly to make it available in real time. Configured and deployed the services, streams in the kubernetes cluster for high availability and fault tolerance. Deal Preparation ML Pipeline: Designed and developed web API's which can provide machine learning model information to data scientists, analysts and managers interactively. Developed the database views which can help as input to ML training and prediction pipelines. Identified the feature importance using machine learning models like XGB, LGB, random forest. Experimented, evaluated the model with different regression algorithms and autoML. Implemented training, prediction processes and scheduled as part kubernetes cron jobs to get the latest predictions and store them in redshift tables. Deployed the entire ML pipeline within kubernetes clusters.

Software Engineer (Big Data & Analytics Practice)

Innominds Software Private Limited

January 2015 - January 2020

Project: Automation platform for security alarms (Client: US-based startup) Project summary: Security teams struggle with an overwhelming number of alarms while still relying on manual assessment and response processes. Typically a security operator is interfacing with 3-4 applications (running on separate monitors) which adds to the operator's challenges. To avoid the false alarm notification we are building an intelligent alarm automation platform. Roles & Responsibilities: Provided multiple API's to train, run and get the inferences of the model interactively. Extracted the features to build the models using descriptive analytics. Experimented, evaluated the model with different regression algorithms and autoML. Deployed and Productionised the regression and text classification model in the Azure cloud. Project: Cognitive Adverse Event Intake and Processing (Client: Deloitte) Project summary: Pharmacovigilance (PV) organisations face a growing volume of adverse event (AE) cases, but today's manual processes can be time-consuming and costly. We designed a way to automate AE processing to help reduce costs and uncover more insights that can improve product safety. We have built an intelligent system that will use cognitive analytics for Recommendations and Signal Detection. Role & Responsibilities: Implemented and integrated all data-related Flask services to multiple systems like BPM, Java, and RabbitMQ. Developed a data extraction service that extracts the textual information from documents (PDFs). Designed and developed an approach for bounding boxes for textual information using computer vision techniques.

Sai Krishna Bachu