Abhishek Kumar

About

Passionate Data Science student with expertise in analytics and predictive model building. Technical proficiency in Python, R, SQL, and machine learning libraries including sci-kit-learn, TensorFlow, PyTorch, and Keras. Experienced in data interpretation and problem-solving with a strong ability to adapt to new environments and work effectively in teams.

(+91) 7888873321abhishek.k.jalandhar@gmail.com

Work Experience

Insurance Agent

Life Insurance Corporation of India Bettiah, India
Oct 2023 — Current

Lead insurance presentations and community building, collaborative reviewing of insurance policies, and presented policies for education and women's benefits.

Projects

Graph Net

Graph Net logo

Web Crawling, Graph Networks, Knowledge Management

Intelligent Knowledge Graph for Data Science Learning. Designed and developed a specialized knowledge graph platform using web crawling techniques, implementing a log-weighted graph network with token system for concept classification.

View project

ARG Data and Modeling

ARG Data and Modeling logo

Python, TensorFlow, PyTorch, Pandas, NumPy, Matplotlib, Seaborn

No-Code Data Science Platform with drag-and-drop modules for data analysis, preprocessing, and model building. Integrated with TensorFlow, PyTorch, Pandas, and NumPy.

Dynamic R Dataset Analysis and Visualization

Dynamic R Dataset Analysis and Visualization logo

R, Shiny, Statistical Analysis

No-Code Statistical Analysis Tool built in R with interactive features for statistical analysis and data visualization.

Stock Price Prediction

Stock Price Prediction logo

Python, Linear Regression, Yahoo Finance, Sklearn

Stock price prediction model using Linear regression in machine learning.

View project

Stock Analysis and Comparison Tool

Stock Analysis and Comparison Tool logo

Python, BeautifulSoup, yfinance, Streamlit

Website for stock price analysis and comparison between two stocks.

View project

Skills

Education

Research Publications

Investigating the Synergistic Effects of Learning Rate, Optimizer, and Activation Function on Convolutional Neural Network Training

IEEE 5th International Conference on Artificial Intelligence and Signal Processing 2025 Impact Factor: 2.8

Convolutional Neural Networks (CNNs) is changing the various fields, particularly computer vision, demonstrating unparalleled performance in work like image detection and computer vision. However, achieving optimal CNN performance on selection of hyperparameters, which are critical yet not learned during training. Among these, the learning rate, optimizer, and activation function significantly influence a network's convergence speed, training stability, and ultimate generalization capability. While individual studies often focus on optimizing a single hyperparameter, the interplay and synergistic effects between these three crucial components are less comprehensively explored. This research empirically investigates how different combinations of learning rates, optimizers and activation functions collectively impact CNN training dynamics and performance. Through systematic experimentation on benchmark image datasets, we aim to identify optimal configurations and uncover the complex relationships that emerge when these hyperparameters are considered in concert. Our findings reveal that how a parameter change effect the performance of the model and also the dynamic learning rate affect the performance of model.

Technologies: PyTorch, TensorFlow, OpenCV, NumPy, Scikit-learn

Languages

Hindi Native
English C2 - Proficient
German A2 - Basic

Certifications

Contact

Tap to call or email: