• 🚀 Gen AI-Powered Database Interaction System
  Built a Retrieval-Augmented Generation (RAG)-based application using Flask, LLM, and ChromaDB for seamless NLP-driven database interactions with PostgreSQL and MongoDB, enhancing query efficiency by 50%.

• 📸 Transfer Learning for Multi-Class Image Classification
  Developed a multi-class image classifier using transfer learning with models like EfficientNetB0, ResNet, and VGG16, achieving a validation accuracy of 92.57%.

• 💳 Buy Now, Pay Later
  Designed a predictive model for customer adoption of a "Buy Now, Pay Later" feature in a retail app, integrating data-driven insights to boost conversion rates.

• 🔢 OctMnist Classification
  Developed a deep learning classifier on the OctMNIST dataset, leveraging CNN architectures to achieve high accuracy in digit recognition.

• 🚀 SVHN CNN Optimization
  Optimized a Convolutional Neural Network for the SVHN dataset, reducing inference time while maintaining robust performance on real-world image data.

• 🤖 Therapy.ai
  Created an AI-powered chatbot using natural language processing to provide mental health support and therapy recommendations in a user-friendly interface.

• 📈 Stock Price Prediction Using Reinforcement Learning
  Built a reinforcement learning model to predict stock prices and optimize trading strategies, resulting in enhanced decision-making for market investments.

• 🏭 Reinforcement Learning Warehouse Robot
  Implemented RL algorithms for robotic navigation and task planning in a simulated warehouse environment, improving operational efficiency and route optimization.

• 😊 Sentiment Analysis using LSTM
  Developed an LSTM-based model to analyze sentiment from textual data, achieving robust performance in classifying customer reviews and social media posts.

• ⏱ Time Series Forecasting using LSTM
  Designed an LSTM network for time series forecasting, delivering accurate predictions across various domains including finance and demand planning.

• 🖼 VGG16 vs ResNet18 vs ResNeXt
  Conducted a comparative study on popular CNN architectures (VGG16, ResNet18, and ResNeXt) to evaluate performance trade-offs in multi-class image classification tasks.

• 🦠 Rabies Prediction
  Developed a machine learning model to forecast rabies outbreaks using epidemiological data, aiding in proactive public health planning and intervention.

• 💰 AI Model Price Calculator (Streamlit App)
  Built a cost estimation tool for token-based AI model usage including LLaMA, Qwen, DeepSeek, and FLUX models. Designed using Streamlit with interactive tabs, CSV upload support, token counting via tiktoken, and dynamic pricing summaries to support budgeting for GenAI projects.

• Full-Stack Development

• Cloud & DevOps Technologies

• Competitive Programming

✉️ Email: asharan2@buffalo.edu
🔗 LinkedIn: Ashutosh Sharan
💻 GitHub: asharan2buff

• 📝 Publication
  Published "Bike Count Sharing Prediction Using Machine Learning" in Ijaresm.

• 🏅 People’s Choice Award
  Won Lam India Hackathon 2021 for building a VR Lab Assistant App (Flutter + Unity3D), earning a $1500 prize.

• 🎗️ Samarpana Marathon
  Raised $50K for martyred soldiers’ families through a student-led campaign.