Hi, I'm Akash BR

Advanced system combining rule-based processing and LLMs for high-accuracy currency, number, and context validation. Features automatic OCR error correction and context-aware financial interpretations.

Robust system using adapted CNN architectures for plant disease diagnosis. Provides detailed insights into causes and recommends tailored solutions to enhance crop health and yield.

Built using Python, NLTK, TensorFlow, and Keras with deep learning model (128-64-dense) trained with SGD, ReLU, and Softmax for accurate intent classification from JSON dataset.

Full-stack web application where artists can post 3D models and 3D scans. Features complete CRUD operations and advanced search functionality for easy asset discovery.

Innovative system that suggests movies based on user's current mood detected through facial expressions, making it easy to select the perfect movie for any moment.

Multiple projects including Stock Price Forecasting, Spam Email Detection, Sentiment Analysis, Medical Data Visualizer, House Price Prediction, and Speech Recognition System using Python libraries.

Smart assistant built by fine-tuning GPT-2 on 1000+ haircut queries, offering quick, conversational, and accurate answers tailored for a barber shop experience.

Simple yet powerful recommendation system that analyzes movie overviews to help users discover films similar to their favorites using vectorized content matching.

EmailTriageAI is a full-stack project showcasing backend, frontend, and AI integration. Built with Flask and MongoDB, it uses NVIDIA Build API models (Qwen, Gemma, DeepSeek) for smart email classification and search. The system provides a clean dashboard to manage emails, view patterns, and get quick stats — all in one place. Fully developed solo, it demonstrates skills across Python, AI, and frontend design.

A Flask web application that fetches Reddit user data (posts, comments, metadata) via PRAW, processes it with SentenceTransformers embeddings, and stores them in Pinecone for similarity search. NVIDIA's Llama model is then used to generate a detailed persona JSON (traits, interests, communication style). The app includes a clean web interface, debug outputs, and Tailwind CSS styling. Built as part of an ML internship assignment to showcase skills in NLP, vector databases, and AI integration.

Mosquito Sentinel is a Streamlit-based AI application that identifies 38 mosquito species from real-time audio input. Built with PyTorch and trained on the HumBugDB dataset with environmental noise augmentation, it delivers robust detection, species-level disease information, and tailored control measures. The system provides a clean user interface, runs in real-time on CPU, and sets the foundation for geolocation-based outbreak mapping and public health monitoring. Developed to tackle mosquito-borne diseases such as malaria, dengue, and Zika with accessible technology for communities, researchers, and health officials.

CampaignIO is a modern influencer discovery and outreach system built with Next.js (frontend), Flask (backend), Node.js + Puppeteer (async discovery worker), and MongoDB Atlas. It streams influencer recommendations in real-time, maintains search history, and offers a full workspace with Timeline, Notes, Events, Tasks, Deals, Campaigns, Messages, History, and Todo. The system is designed to feel fast and resilient — results appear incrementally while discovery continues in the background. Deployed via Docker on Render, it integrates Instagram Graph API and Google OAuth for seamless use.