To predict the tool breakages in a cold forming machinery line for a Tier 2 Automobile Supplier by collecting the relevant and fruitful data from various IT systems such as SAP ERP, PP, WM etc and colle cting sensors data from the Edge Devices.

• Conceptualized a Semantic Business Rule Engine for Edge devices so that huge chunk of processing can be done on edge device itself and creating only fruitful data for analytics purpose
• Implemented the rules to harmonize the data in data lake using ETL tools such as Hive and Kafka
• Worked on generating the ranking of various input parameters for given business use case by utilizing the algorithms such as Boruta, RFE, BE etc
• Helped plant users to have clea r visibility about the scrap coming out of machine and further giving them recommendations to reduce the same

• Gathered requirements of retail domain experts and set the expectations in terms of FP, FN, TP, TN
• Optimized SQL queries using complex joins to reduce data retrieval time
• Implemented the data pipeline for computing KPIs from consumer data base, environmental sensors. Computing device and log data
• Implemented data preparation and parallelized mass training of forecast models (multiple 100k models trained)
• Built custom dashboard for quick assessment by controllers

To predict the quality status of Alumina’s (AL2O3) Specific Surface Area (SSA) for upcoming batches and taking corrective action on input process parameters to prevent and reduce the number of faults (slides) and anomalies.

• Created data pipelines and workflows to extract data from machines like KILN, Combin Rotators, Cooling Tower, Multiclone and Polyclone Precipitators, Pan Filter
• Implemented Mahalanobis Distance function based Data Cleanup Algorithms given the data distribution
• Selected decision tree based algorithms by analyzing the distribution of given dataset and further Benchmarked Scikit based RF, XGB and H2O based XGB algorithms to come up with hi gh accuracy deployable model
• Built a 2 hours in advance prediction of SSA quality attributes and ranking of input process parameters in the order in which they influence a given quality attribute 

Predicting the quality status of aluminium coil coming from TCF Casting Machine for upcoming batches and taking corrective action on input process parameters to reduce the number of defective pieces

• Gathered requirements of metal domain experts and collecting the data from machines like twin chamber furnace, caster, m illing, VDU units, and further from quality lab test data
• Researched Bayesian Optimization for the given data and setup open source implementations thereof
• Created GLM models to handle scaling of various process parameters data
• Built web app for visualizing experiment data and surrogate function as well as proposing new formulations to test next

• Gathered requirements of printing press domain experts and set the expectations in terms of FP, FN, TP, TN
• Developed algorithms to identify anomalies and investigate patterns in the existing parameters like motor speed, voltage etc.
• Developed AI engine using machine learning algorithms for predictive analytics with evaluation and risk matrix
• Built customized dashboard and mobile based application for quick and easy access to the data and KPIs

To reduce inventory cost by predicting whether OEM needs a backup battery in the truck or not.

• Collecting requirements from asset management company which gives trucks on a rental model basis
• Created multivariate regression model to predict the health of battery by using parameters such as distance covered, fuel consumption in eco/start mode, battery current, voltage etc.
• Deployed the model in real time on OEM dashboard showing the health of bat tery of all trucks in the fleet

To optimize long term and intraday contracts by predicting energy consumption and generation uncertainties.

• Gathered requirements of energy domain experts and set the expectations in terms of FP, FN, TP, TN
• Creat ed data pipelines and workflows to extract data from SCADA systems, grid operator, log files, market and relevant indexes
• Developed AI/ML engine to predict energy supply and demand uncertainties
• Built web based application for visualization of relevan t KPIs inbuilt with all the simulation and prediction models for different energy trading positions