Kshitij Singh
Development
Uttar Pradesh, India
Skills
Data Science
About
KSHITIJ SINGH's skills align with IT R&D Professionals (Information and Communication Technology). KSHITIJ also has skills associated with System Developers and Analysts (Information and Communication Technology). KSHITIJ SINGH appears to be a low-to-mid level candidate, with 5 years of experience.
View more
Work Experience
Data Scientist
Pivotchain Solutions
August 2021 - Present
- Primary responsibilities included Data Science model design & development, Data Engineering & transformation, Creating, Enhancing & maintaining ML models. • Contributed extensively to a product RavenAI, that monitors the ATMs(more than 1000) and has alert raising mechanism for unusual activities. • Developed a Retrieval Augmented Generation (RAG) application for context-aware question answering capabilities. • Utilized CNN to build custom image classification model to help client monitor the site without human pesence. • Lead a team of 4 data scientists to deliver an end to end product including development and deployment to the defence agency of India. • Built custom trained multilingual OCR(PARSeq) for Number Plate Detection. • Built an Adversarial model to generate synthetic data to reduce class imbalance problem for object detection model, eventually increasing the accuracy by 5% for particular classes. • Developed an automated data pipeline for video data to extract useful frames, reducing processing time from several hours to just a few minutes without human intervention. • Automated annotations for image dataset using Nuclio and reduced the data preparation time by 40%. • Expertly handled data engineering of over 800,000 images, balancing classes and splitting datasets for improved YOLOv5 performance - higher mAP, superior accuracy, and model generalization. • Created real time data visualization Analytics dashboard to keep track of more than 1000 AI systems all over India.
Machine Learning engineer
Proxykhel Pvt Ltd
June 2020 - July 2021
- Engineered a User Churn Risk Model for a user base exceeding 150,000, aiming to gain insights into user behavior dynamics and mitigate churn rates effectively. Employed XGBoost, to develop predictive models that accurately forecast user churn probabilities and inform strategic decision-making. • Crafted a competitive gaming algorithm tailored for user segments ranging from 10,000 to 100,000, simulating diverse gaming scenarios to optimize engagement and retention strategies. • Applied a diverse array of statistical techniques to validate simulation outputs, culminating in the development of a comprehensive scoring algorithm.
Research Assistant
Defence Institute of Advanced Technology
June 2019 - June 2020
- Stock Market Prediction Using LSTM: Time Series Analysis, Feature Engineering, Model Building, Hyperparameter Tuning • Election Data Analysis: Data Collection, Data Preprocessing, Data Visualization, Natural Language Processing
Education
Defence Institute of Advanced Technology
Master of Technology
