Skip to content

Kumaresh1110/portfolio

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

68 Commits
 
 
 
 
 
 

Repository files navigation

Kumaresh Vijayakumar

Email: [email protected]

Phone Number: +1(872)-233-5193

LinkedIn: Kumaresh Vijayakumar

Education

Masters in Computer Science - DePaul University

Work Experience

Programmer - Applied Philosophy of Science.

Chicago Illinois.

As part of a comprehensive NLP project, I led the development of an environment in Google Colab, utilizing pre-trained transformer models to process and analyze textual data. The project centered on building Named Entity Recognition (NER) functionality with BERT and initial explorations in answer generation with GPT-2.

Project Highlights:

Google Colab Setup & Environment Configuration:

Installed essential libraries like transformers, torch, and datasets, preparing Colab for NLP processing. Integrated Google Drive for easy dataset access and streamlined data handling.

Data Preparation & Preprocessing:

Loaded and preprocessed datasets using pandas and the datasets library, transforming raw CSV files into training-ready formats. Assisted in data creation by structuring and organizing AP questions for model training.

BERT for Named Entity Recognition (NER):

Implemented pre-trained BERT tokenizer and model for token classification tasks. Aligned tokenized inputs with their respective labels to ensure accurate and efficient entity extraction. Fine-tuned the BERT model, configuring training parameters using Hugging Face's Trainer module for optimal results.

Code Notebook Development:

Created an end-to-end Python notebook in Google Colab, automating data preprocessing and BERT fine-tuning for efficient project execution.

Model Optimization:

Began with GPT-2 fine-tuning for answer generation, but transitioned to BERT for more task-specific accuracy in NER, demonstrating adaptability to project requirements.

Collaborative Efforts:

Engaged regularly with the team and professor, discussing model optimization, data creation, and overall project goals. Integrated feedback into the project, ensuring alignment with both academic and technical objectives.

Key Technologies & Tools:

Libraries: transformers, torch, datasets, pandas.

Models: BERT (for NER), GPT-2 (for answer generation).

Tools: Google Colab, Jupyter Notebooks, Google Drive.

Impact: Improved accuracy of entity recognition tasks using fine-tuned BERT models, laying the groundwork for future NLP applications. Showcased the ability to adapt and fine-tune state-of-the-art models for specific text processing tasks. Gained invaluable hands-on experience in model training, data manipulation, and problem-solving in real-world scenarios.

Enterprise Analytics Intern - Midland States Bank, Rockford Illinois.

Rockford Illinois.

During my internship at Midland States Bank, I worked closely with the Enterprise Analytics team under the supervision of Megan Lloyd. My primary responsibilities included creating and modifying Cognos reports and Power BI dashboards to support data-driven decision-making processes across various business departments.

Key Contributions: Power BI Dashboards and Cognos Reports

Developed and optimized 35 Cognos reports and 2 Power BI dashboards, providing critical insights that improved decision-making for business stakeholders. Streamlined data visualization processes, reducing time for stakeholders to access and interpret key data points.

Reporting Request Management:

Managed and tracked reporting requests through the Remedy ticketing system, ensuring timely resolution of tickets and maintaining an organized workflow.

Stakeholder Collaboration:

Actively engaged with business stakeholders to gather requirements and present solutions, ensuring the dashboards and reports met their specific needs.

Team Collaboration:

Participated in daily stand-ups, grooming sessions, and code reviews, providing valuable feedback and staying up-to-date on ongoing projects within the team.

Skills Acquired:

Data Visualization: Gained hands-on experience creating visually appealing and informative data visualizations using both Cognos and Power BI.

Technical Proficiency: Improved technical skills in data manipulation, data modeling, and dashboard development, enhancing the bank’s analytics capabilities.

Communication & Collaboration: Refined communication skills through regular interaction with stakeholders and team members, aligning technical solutions with business needs.

Time Management: Developed effective time management skills by prioritizing multiple reporting requests and meeting deadlines consistently.

Impact: My work during the internship was well-received, with business stakeholders acknowledging the accuracy and usefulness of the insights I provided. The skills I gained in data visualization, technical reporting, and collaboration will serve as a strong foundation for my future career in data analytics.

Selected Projects

Enhancing Object Tracking in Game Clips Through Image Processing Techniques.

• Applied image processing techniques such as Gaussian Blur, Canny edge detection, and Salt and Pepper noise to enhance object tracking in game clips. Improved the accuracy and performance of tracking algorithms by preprocessing and refining video frames. Conducted quantitative analysis to measure the impact of each technique on tracking precision and stability.

• Collaborated with a team to integrate these techniques into a real-time tracking system for live game clips.

Unpaired Image to Image Translation using cycleGAN.

• Utilized cycleGAN for unpaired image-to-image translation, enabling the transformation of images from one domain to another without paired examples. Enhanced the quality of translated images through extensive parameter tuning and model training. Conducted comparative studies with other GAN-based models to validate the effectiveness of cycleGAN.

• Presented findings in a seminar, showcasing the practical applications of unpaired image translation in various fields.

Enhancing Object Tracking in Game Clips Through Image Analysis Techniques.

• Implemented image analysis techniques including Segmentation, Gaussian Filtering, and Haar Image Compression to improve object tracking in game clips. Developed custom algorithms for segmenting moving objects in complex backgrounds. • Reduced computational load and increased efficiency of tracking systems through optimized image compression methods.

ASL Alphabet Recognition: A Deep Learning Approach.

• Developed a deep learning model using ResNetv2 for recognizing American Sign Language (ASL) alphabets. Collected and preprocessed a large dataset of ASL images to train the model. Achieved high accuracy in alphabet recognition through rigorous model training and validation. Created a user-friendly application to demonstrate the model’s capabilities, facilitating real-time ASL recognition.

Certifications

Udemy:

Microsoft Power BI Desktop for Business Intelligence (2023).

Learning Python.

IBM Cognos.

The Ultimate MySQL Bootcamp.

Python Programming - Multithreading, OOP, NumPy and Pandas.

Statistics for Data Science and Business Analysis.

Writing production-ready ETL pipelines in Python / Pandas.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages