Himadri Sikhar Khargharia brings 14 years of experience in data analysis and software engineering, with a strong background in designing and developing enterprise-grade applications. His expertise spans data analysis and machine learning, particularly in the domains of online retail and indoor localization. Over the years, he has made significant contributions across various fields, including e-Learning, supply chain (order management), IoT (personalization), indoor positioning, workforce management, and planning and scheduling software. Notably, Himadri was awarded the Best Application Paper Award at the BCS AI Conference in 2019, recognizing his contributions to applied AI solutions.
Resource Optimization
In this project, multi-objective optimization was implemented to increase capacity, reduce shrinkage, and minimize contingency and overtime, leading to efficient utilization of resource capital. The project also focused on enhancing forecasting accuracy for work stack demand by tuning the hyperparameters of prediction models with evolutionary algorithms and leveraging temporal patterns in sequence prediction to reduce the number of models needed for accurate forecasts. Additionally, a platform was developed for British Telecom to facilitate data analysis visualization, workforce planning, scheduling, and demand prediction. Built on a microservices architecture using Spring Boot, the platform's frontend was created with Angular 6 and Node.js, and it incorporated ElasticSearch and Logstash for data indexing and synchronization.
IoT Personalization
In this project, privacy-preserving, edge-based personalization was achieved using contextual information from an indoor positioning system (IPS) in a large-scale retail environment. The project included detection and generation of hierarchical relationships among points of interest within the IPS, and statistical machine learning methods were employed to develop classification models for detecting points of interest using data from WiFi, BLE, magnetometer, and IMU sensors. Bayesian networks and probabilistic graphical models were used to recommend user preferences based on past behavioral patterns. Additionally, the project involved developing an IoT middleware stack called Bezirk, an Android-based user profiling server, and an indoor positioning system compatible with Android, iOS, and cloud-based platforms.
Distributed Order Management (Supply Chain)
In this project, scheduling and allocation of SKUs from warehouses were managed as part of Distributed Order Management using Manhattan's product, including tasks such as requirement gathering, module design, development, and providing support during testing and Go-Lives.
ELearning Software For Autistic Children
In this project, ontology-based semantic inferencing, along with ANN and decision tree-based learning, were used to deliver personalized digital content tailored for children with autism. The project also involved the development of a server-based, multi-modal eLearning software specifically designed to support the educational needs of autistic children.
