Aditya Kumar Barik Email & Phone Number

Bengaluru, Karnataka, India

1. Engineered a Table Processing Model to extract tabular data from scanned images containing clinical data, utilizing Microsoft's DETR-based table detection and structure recognition models from Hugging Face. Attained 90% accuracy in table detection and 66% in structure recognition, improving data extraction efficiency by 20% and reducing OCR-related complexities by nearly 50%.2. Tackled a complex checkbox classification challenge involving noisy image data by deploying a shallow Convolutional Neural Network (CNN) model, based on a research study. Attained 89% accuracy despite noisy training data.3. Built a comprehensive pipeline for accurate information extraction from noisy image data based on client requirements, achieving 70% accuracy. Introduced a novel datatype class for streamlined output.4. Amplified project potential by integrating advanced Large Language Models (LLMs) such as BERT and GPT-2 from Hugging Face, leveraging insights from leading research. This significantly boosted project performance.5. Spearheaded a multi-label text classification project, leveraging Logistic Regression, Random Forest, and XGBoost models. Achieved an impressive 94.12% increase in Macro F1-Score through data preprocessing and optimization.6. Standardized model development process using the MLflow framework in team of 5, enhancing reproducibility and transparency; leveraged the Optuna for hyperparameter tuning, achieving 15% faster convergence and boosting model performance metrics by 8%.7. Implemented MLOps practices to streamline and automate the deployment, monitoring and management of machine learning models, ensuring continuous integration and deployment (CI/CD) across various projects.

Gurugram, Haryana, India

1. Automated sales predictions for a global conglomerate, handling 70+ explanatory variables with PCA and Ridge Regression. Addressed confounding variables and causality challenges to enhance accuracy.2. Developed a vehicle routing optimizer for an Indian chemical company, accommodating more than 10 constraints, by harnessing the ortools package in Python and Google API.3. Executed comprehensive business process automation for a global BFSI firm using Pandas; developed a Python Tkinter GUI, cutting processing time by 80% and enhancing operational efficiency by 60%.

Kolkata, West Bengal, India

1. Pioneered a highly efficient Resume Parser application, combining NER, NLP, RegEx, Face Detection and other advanced techniques. Seamlessly integrated with MongoDB for data storage and retrieval, reduced manual resume processing time by 40% and enhanced recruitment efficiency significantly.2. Conducted exploratory data analysis of patient records, using Multiple Logistic Regression to predict diseases with 67% accuracy and uncover health trends and patterns.3. Implemented a robust backend REST API server using Flask for efficient patient data processing and seamless ML algorithm execution.4. Crafted a customized chatbot that analyzed text context and sentiments, enhancing user interactions by 30% and streamlining information retrieval, reducing response times by 40%.

Kolkata

1. Developed a Time Series (SARIMA) model using NSRDB data from 90 meteorological stations across India and predicted 7-day ahead Global Horizontal Irradiance (GHI) with an impressive 85% accuracy. Presented at ICDMAI 2020 and published in Springer.2. Investigated the impact of the COVID-19 pandemic on the aviation industry's stock market through Time Series Segmentation in R. Explored log-returns during various COVID-related events. Published in a special JISA issue in June 2020.3. Performed a Spatio-Temporal Analysis of GHI, unraveling the relationship between diverse meteorological variables and solar irradiance.

Data Science

Master Of Science - Msc, Statistics And Informatics, 81.55%

Bachelor Of Science - Bsc, Statistics, 57%

Higher Secondary, 76%

Secondary, 87.4%