Kratika Khare

Product Management

Collexion (Jan 2022- Present)

• Testing and retraining of all models to validate performance before deployment. Validated 5000+ user test cases (random samples) to verify model worked as designed
• Led team of 4 and managed cross-functional teams and stakeholders like data collection, data scientist interns, model deployment team and oversaw product development from inception to implementation.
• Proactively suggest ways and feature to incorporate in product to improve client conversion. Identified products that was under performing, and recommended improvements that can touch 90% accuracy.

CAX Community (March 2019 - Present)

• Design competitive platform description for various Machine learning / Deep learning algorithms.
• Model validation of top models in contest on test data for Over-fit verification. Selection of robust model from contest which is production grade in real environment.
• Concise report preparation for clients with detailed insight reports and interactive dashboards to showcase optimization test learning.

Key Projects
Back ground removal (2020 - Present)

• Fashion/Product Background removal : Applied image matting technique to input trimaps along with image as input to extract delicate segments like hair strands during foreground separation in background removal technique.
• Automobiles: Use pretrained weights from U2Net architecture. Best checkpoints are saved and model trained on custom dataset.
• Furniture Background removal: High resolution images are collected and pretrained weights from IS-Net architure used to segment sensitive cases like light color object in light background, complex room setup furniture, etc.

Shadow/Refection Generation (June 2022 - Present)
Creating shadow for fashion images using image processing techniques. Adds circular shadow at bottom of product to make it e commerce ready.

Freelime Prediction (2019)
Applied Random-forest /XGBoost and certain ensemble approaches to get idea of cement produced a head of time which will save power/money/time

Mortgage Funding (2019)
Develop a model using XG Boost/Light GBM to predict, given mortgage application information, whether mortgage will be Funded or Not Funded and obtained 0.75 Macro average F1 score with 90 percent precision

Credit Default Model (2020)
Developed credit default scoring model with data with less variability to predict probability of defaulters( for low income or less or no credit score people) using Variable selection by WOE and IV to break base line accuracy there by improving recall of defaulters by significant amou

KKC / SOMPO / SIG (2021)

• Helps client in requirements/USDM analysis, timely comments on xd design to give appropriate feedback to clients and helped them in user friendly UI designs creation.
• Created UI wireframes to upgrade existing machine learning platform.

AiryQonnect (2021)

• Created UI wire-frames to create sensor board to add multiple sensor devices in restaurants to check its climate for customers
• Worked with testing team to ensure proper functioning of app as per client requirements.

• Web scraping of real estate data where more than 50000 property links are extracted along with their parameters like furniture status, dimension, etc.

• Assisted 120 students and applied innovative teaching methods to encourage student learning objectives.
• Designed Curriculum of Microprocessor which is project based and gives hands on experience to students

• Handled undergraduate course classes and took lecture in TV and Radar