Ashok Kumar

2022-Dec - 2024-Sep

• L&T Construction (MAHSR) – Computer vision-based production factory monitoring to produce production insights. Setting on prem servers and inference pipelines. Improved production cycle and bottleneck steps
• L&T Heavy Engineering - In house Design and Developing laser based portable 3d profiling device. Integrated geometry-based defect monitoring for weld defects. Reduced NDT by 4 hrs
• POC - Contact pattern defect classification using object detection models. Proposed Laser 3d profile creation for NDT
• Ltimindtree - Smart Parking (Smart Spaces Product) solution using ANPR and QR to monitor ingress/egress.
• Time Series Forecasting for Ltimindtree ESG reporting
• PoC - Navistar - Big data analysis on vehicle OBD sensor data Providing insights for vehicle segments, age, driver profile
• POC - AVIS – Car dent detection using segmentation models
• Azure Digital Twin generation for Realtime monitoring rented vehicle and dent changes.
• POC - GE Healthcare – Machine Log Analytics. GEN AI based logs analysis and troubleshooting.
• POC - Structural Health Monitoring using accelerometer data RUL and predictive maintenance
• Synthetic data generation for back testing anomaly detection models

Hardware: Nvidia Jetson Nano/Xavier, Raspberrypi 4b, Intel depth sense d435i, Line Laser, Lucid Vision Industrial Camera, Rapid Prototyping (3d Printer), Arduino, ADXL345

Technology: Python, C++, Arm64/32, Matlab, Free Cad, Creality slicer/Cura, Label Studio, QT Creator, Paraview

Libraries: Tensorflow, Pytorch, YoloV8, openvino, Opencv, GAN, Style transfer, SDV, QT6, Fast API, Streamlit,, Flask

2020-Aug to 2022-Dec

• Developed **active learning** for model training and data labelling
• Jaguar and Landrover – Consumer data analytics. Achieved 6% lead conversion utilizing logit model. Managing ML Model Pipelines using GCP Matillion. Managing SVCRM tables with big query.
• Code migration from R to Python.
• Customer churn prediction using survival models
• P&G - Customer sales data analysis and building recommendation engine. Awarded for best delivery project
• POC – Process optimizing using Linear/Non-Linear programming (Staff Scheduling, Supply Chain)
• POC - Benchmarking ML Tools and Platforms (Samsung sds, Knime, H2O, Azureml, datarobot).

Technology: Python, R, GCP, Matillion, SVCRM, Brightics-Al (Samsung), KNIME, Pyspark-ML, Azure ML, Vector-Al

Libraries: Pyomo, Gekko, Gurobi, Excel Solver, Scikit-Learn