MLOps Platform Development: Architected an MLOps platform supporting ML lifecycle (training, deployment, monitoring) with Kubernetes, Docker, and Terraform, reducing deployment time by 30%.ML Infrastructure Design: Built feature store and model repository using MLflow and Kubeflow, enabling efficient development, tracking, and versioning across complex ML projects.CI/CD Pipeline Automation: Developed automated CI/CD pipelines using Jenkins and GitHub Actions for ML model testing and deployment, establishing robust testing and validation processes.LLM Integration: Integrated LLM frameworks (LangChain, LLaMA) for customer service automation, achieving 3-second response times through optimized model serving architecture.Model Observability: Implemented comprehensive model observability using Datadog and Grafana, tracking data drift, performance metrics, and system health with 40% reduced downtime.

Inference System Architecture: Built a distributed ML inference system using TensorFlow-Serving and Triton, efficiently scaling to process over 1M+ daily inferences across healthcare modalities, deployed on AWS EKS.Data Pipeline Development: Designed and developed a scalable, fault-tolerant data processing pipeline capable of handling terabytes of structured and unstructured healthcare data using Apache Spark and Kafka streams.Security & Compliance: : Established a secure, GDPR-compliant ML infrastructure on AWS, implementingrobust data handling protocols and access controls for sensitive patient information across the organization.ML Lifecycle Management: Led the deployment of Kubeflow and MLflow for experiment tracking, model versioning, and automated deployment across a multi-tenant architecture, enhancing operational efficiency.Team Leadership: Managed a 14-member ML engineering team, implementing best practices for model development, deployment, and compliance with strict healthcare regulations, ensuring timely project delivery.

MLRT: Led the development of MLRT, ML inferencing engine designed to provide scalable, effective inference for ML models such as CatBoost, XgBoost and Scikit-Learn using Python, NumPy, Pandas, microservice API design and API security to enable flexible integration with CI/CD/DevOps ML pipelines and multi-tenant architectures.Onboarding and Inferencing: Collaborated with a cross-functional team to onboard and perform inferencing of ML/DL models using Docker containers and deploying them to AWS EKS with S3 for storage. The models were used for various healthcare modalities like MRI, CT and PET, and data storage, management, and processing was handled using EFS, RDS, SQS, and EC2. Kubernetes was utilized for managing the containers and scaling the application.

ML Service Architecture: Developed scalable microservices architecture for ML model deployment using Python, SOLID principles, and containerization, enabling efficient scaling and streamlined integration.Data Analysis Tools: Built a real-time data processing pipeline for analyzing aircraft telemetry data, handlinghigh-velocity streaming data with optimized performance and actionable insights for engineering.DevOps Implementation: Implemented automated testing frameworks and CI/CD pipelines, ensuring reliable and consistent deployment of ML services across environments, enhancing operational efficiency and minimizing errors.Agile Leadership: Served as Scrum Master for an 8-member team, leading sprint planning, stand-ups, and retrospectives, achieving a 25% increase in project velocity and improving team alignment.

◦ Vibration Lab: Worked on a research study under the guidance of Prof. AK Darpe that examined the propagation of sound waves through stethoscopes when placed on human hearts. Used specialised tools to collect data and cutting-edge signal processing techniques to analyse, identify trends and patterns in the sound waves.◦ Bearing Failures: Developed a technique to identify bearing failures utilising Arduino microcontrollers through online monitoring of the vibration patterns of CNC machines. A green light indicates that all of the bearings are in good condition, whereas a red light indicates that one or more bearings need to be changed. This is a reliable and efficient technique for detecting and eliminating potential issues in industrial machinery

Designed the websites http://almafiesta.com/2014/journey/ , http://almafiesta.com/ in the capacity of Core Head of the Web Team, Alma Fiesta.

◦ Investigations on simultaneously developing flows in micro-channels: Developed models in COMSOL 5.0 to enhance heat transfer in channels of different shapes and curvatures by optimising Reynolds number, Stanton number and velocity, validated over 200 simulations in COMSOL, developed a MATLAB code using “SIMPLE Algorithm” for a rectangular channel to calculate Nusselt number.

Promulgated fest through ground publicity.