I'm a Data Science and Artificial Intelligence graduate from IIIT Dharwad. Most of my work lives somewhere between healthcare data and ML pipelines — I've spent time building NLP chatbots for clinical domains, classifying facial skin types with CNNs, and co-authoring a paper that got published in a Scopus-indexed journal.

I like problems that are messy in the real world: noisy data, domain-specific language, limited labels. That's where the interesting engineering happens.

B.Tech in Data Science and Artificial Intelligence · IIIT Dharwad · 2021–2025 · CGPA: 7.13
Intermediate · Aditya IIT Junior College, Kakinada · 2019–2021 · 93.6%

Impact of Internet Derived Information Obstruction Treatment (IDIOT) Syndrome — A Student Cohort Study
Tuijin Jishu / Journal of Propulsion Technology · Scopus Indexed · Vol. 47, 2026
Chava Srinivasa Sai, Mandapalli Ruthvik, Ramesh Athe

A retrieval-based chatbot for answering glaucoma-related clinical questions. Uses BioBERT and BioWordVec in parallel — returns two answers with confidence scores so you can compare semantic match quality across both embedding spaces. Dataset: 1,679 curated Q&A pairs. Similarity computed via cosine distance on precomputed embeddings (serialized for fast inference).

Python HuggingFace Transformers Gensim BioBERT BioWordVec Scikit-learn

End-to-end ML pipeline for classifying skin as oily, dry, or normal from facial images. Built a hybrid feature extraction approach — HOG descriptors for texture + MobileNetV2 for deep features — then ran classification through ResNet50 and VGG16. Preprocessing chain: Gaussian Blur → Bilateral Filtering → CLAHE.

ResNet50: 84.16% avg accuracy (±4.58) across 5-fold CV. Dataset was 166 manually curated images, ages 9–40, captured at 50MP.

Python TensorFlow/Keras OpenCV ResNet50 VGG16 MobileNetV2 HOG Scikit-learn

ML-driven cohort study on "Internet Derived Information Obstruction Treatment" syndrome — a pattern of excessive health-related searching that drives anxiety. Surveyed 450 technical students, built a Random Forest classifier (68% accuracy) and Multi-linear Regression risk scorer (R² = 0.62, MSE = 1.35). Feature engineering via TF-IDF + StandardScaler normalization. Statistical validation: Cronbach's Alpha 0.83–0.87.

Published in Tuijin Jishu / Journal of Propulsion Technology (Scopus Indexed), Vol. 47, 2026.

Python Pandas NumPy Scikit-learn Random Forest TF-IDF Matplotlib

A responsive LinkedIn-inspired front-end, built from scratch. Covers navigation, profile panels, content feeds, and dynamic DOM interactions. No frameworks — just vanilla HTML, CSS, and JS.

HTML CSS JavaScript DOM Manipulation Responsive Design

Languages: Python · C · C++
ML/DL: Random Forest · ResNet50 · VGG16 · MobileNetV2 · CNN architectures · cross-validation
NLP: BioBERT · BioWordVec · Word2Vec · TF-IDF · cosine similarity · Transformers
Data: Pandas · NumPy · Matplotlib
Frameworks: TensorFlow/Keras · PyTorch
Web: HTML · CSS · JavaScript · responsive design
CS Fundamentals: DSA · OOPs · DBMS