dharunkumar

About Me

I am Dharun Kumar, a dedicated and research-driven Artificial Intelligence enthusiast with a strong inclination toward open-source contributions and real-world problem-solving. My interests lie in the intersection of AI, sustainability, and social impact. I actively pursue hands-on projects and interdisciplinary applications to expand my knowledge and contribute meaningfully to the AI community.

Tech Skills

Existing:

• Programming & Libraries: Python, NumPy, Pandas, Scikit-learn, Matplotlib
• Deep Learning: TensorFlow, Keras, PyTorch
• Natural Language Processing: SpaCy, NLTK, Transformers
• Computer Vision: OpenCV
• Models & Techniques: Linear/Logistic Regression, CNNs, LSTMs, Anomaly Detection
• Tools: Git, GitHub, REST APIs

Planning to Learn:

• Retrieval-Augmented Generation (RAG)
• Parameter-efficient fine-tuning (LoRA, QLoRA)
• Open-source LLMs (LLaMA, Phi-3, Mistral)
• Multi-modal AI (text, image, audio integration)
• LangChain and autonomous AI agents
• TinyML and deployment of LLMs on edge devices

Projects I Worked On

Research Contributions:

1. Evaluation of AI-Powered Spelling and Grammar Correction Tools
   Developed a lightweight NLP-based correction system using transformer models and rule-based logic.
   Published on Zenodo: Link

2. Coral Health Monitoring and Bleaching Detection
   Designed a MobileNetV2-based classifier to detect coral bleaching from underwater images, aiding marine conservation research.
   Published on Zenodo: Link

Academic & Independent Projects:

1. Weather Trend Prediction Using Linear Regression
   Integrated OpenWeatherMap API to develop a 7-day temperature forecasting tool using linear regression.
   GitHub: Repository

2. Diabetes Prediction with Logistic Regression
   Created a predictive model using the Pima Indians dataset, applying feature engineering, EDA, and classification metrics.
   GitHub: Repository

3. Domino’s Stock Price Prediction using LSTM with Anomaly Detection
   Implemented a deep learning model with residual-based anomaly detection to forecast stock price fluctuations.
   GitHub: Repository

4. Waste Image Classification using ResNet18
   Built a binary classifier for recyclable vs non-recyclable waste using the TrashNet dataset and PyTorch.
   GitHub: Repository

Certifications

• IBM Deep Learning with PyTorch, Keras, and TensorFlow – Professional Certificate
  Credly Verification

• Generative AI: Foundational Models for NLP and Language Understanding – IBM
  Coursera Verification

• Natural Language Processing in Microsoft Azure
  Coursera Verification

• GenAI-Powered Data Analytics Job Simulation – TATA
  Forage Verification

Aspirations

I aim to become a Data Scientist and AI Engineer with a strong focus on developing scalable, ethical, and impactful AI solutions. My long-term vision includes contributing to national and global public welfare initiatives through the responsible application of Artificial Intelligence in areas like water quality, climate resilience, and healthcare access.