Vipanshu Kumar Suman

Created an AI-driven web application that analyses food labels from a single image upload, delivering ingredient breakdowns, nutrition facts, and allergen alerts. Integrated Google Gemini 1.5 Flash for advanced OCR and structured JSON extraction across multiple products. Developed a personalised “Health Score” engine that flags unhealthy or misleading items based on user profiles. The tool uses Streamlit for a responsive UI and leverages Pandas and Pillow for real-time insights, empowering healthier, data-informed food decisions.

Created a web-based program in AI that can automatically make descriptive photo captions based on the BLIP (Bootstrapping Language-Image Pre-training) model. Made use of Hugging Face Transformers for highly accurate captioning as well as PIL for quick processing of an image. Also added a Gradio interface for live, interactive processing of the captions as well as an export function for writing the results out in a CSV file for added user-friendly functionality, providing speed, accuracy, and usability in a single program.

Designed an IoT real-time weather tracking system with an ESP8266-based microcontroller and DHT11/BMP180 sensors for tracking temperature, humidity, and air pressure. Ensured easy data transmission with ThingSpeak API over Wi-Fi for cloud-based visual interpretation and historical trend tracking. Showcased a 25% greater data reliability with sensor calibration and multi condition testing.

Created a smart resume builder that produces professionally styled, LaTeX-based PDF resumes with AI-optimized content. Employed Jinja2 for runtime templating and presented user data in a short, recruiter-friendly layout. Intuitive Streamlit frontend offers live customization, with Puter.JS serving fast, client-side rendering for near-immediate creation of a PDF streamlining the entire resume-building experience.

The IMDB Movie Review Sentiment Analysis project utilizes machine learning to classify movie reviews as either positive or negative. It employs an LSTM (Long Short-Term Memory) network for sequence-based sentiment prediction, with text data preprocessed through tokenization and padding to ensure model compatibility. The model is trained on a labeled IMDB dataset to perform accurate binary classification. Real-time predictions are delivered via a user-friendly Gradio interface, enabling seamless interaction and instant sentiment analysis.