Open to opportunities

Sriharsha Challangi

@sriharshachallangi

Message

I am a data scientist specializing in machine learning, deep learning, and model interpretability.

United States

Message

What I'm looking for

I am seeking an entry-level data science role building ML/DL models with interpretable outputs, real-world deployment, and collaborative, growth-oriented teams.

I am a Computer Science graduate focused on machine learning, deep learning, and interpretable models, with a strong foundation in statistics and data analysis. I build production-oriented solutions that emphasize transparency and robust performance.

I have developed end-to-end projects including a casting defect detector (99.86% test accuracy), a plant classifier (97% accuracy), and a fake news detector (98% accuracy on 44,000+ articles), plus a multilingual real-time translator using transformer models. I regularly use SHAP visualizations, TF‑IDF pipelines, EfficientNetB0 fine-tuning, and Gradio interfaces to make models explainable and accessible.

I work primarily in Python with TensorFlow, PyTorch and a modern data stack (Pandas, NumPy, Scikit‑learn, Matplotlib, Seaborn) and value clean preprocessing, targeted augmentation, and rigorous evaluation. I enjoy collaborating to move ML models from prototype to reliable, user-facing applications.

Experience

Work history, roles, and key accomplishments

Plant Classifier

Personal Projects

Fine-tuned EfficientNetB0 with dropout and global pooling to achieve 97% accuracy on plant image classification, applied targeted data augmentation to improve generalization and deployed a Gradio web interface.

TensorFlow Gradio Python NumPy

Casting Defect Detector

Personal Projects

Developed a PyTorch CNN to automate defect detection in industrial metal casting, achieving 99.86% test accuracy and reducing manual inspection error; added SHAP visualizations for pixel-level interpretability.

PyTorch Convolutional Neural Networks SHAP scikit learn NumPy Computer Vision

Fake News Detector

Personal Projects

Built a logistic regression model on 44,000+ labeled articles using TF-IDF features to achieve 98% classification accuracy and used SHAP to surface token-level drivers of fake vs. real predictions.

Python NLP TF IDF Logistic Regression NLTK SHAP Pandas scikit learn Gradio

Mobile EDA & Price Predictor

Personal Projects

Preprocessed a mobile phone dataset by standardizing textual specs and removing outliers, shifted from polynomial to ensemble models to improve generalization by ~30% and visualized feature impacts.

Python Pandas scikit learn data preprocessing Matplotlib SeaBorn

Multilingual AI Translator

Personal Projects

Developed a real-time multilingual translation app using Facebook's NLLB-200 transformer to support 50+ languages with text and voice I/O, integrating speech recognition and text-to-speech for robust audio workflows.