I’m an AI/ML and Web3 enthusiast passionate about solving complex problems and building innovative solutions. A Versatile Developer Specializing in Backend, AI & ML, and Web3 Development with a passion for Continuous Learning. I enjoy exploring new technologies and creating user-focused, impactful applications.
-8.77/10.0-EFEEE9?style=for-the-badge){kind=icon}
- AI/ML Intern at SmartInternz (Remote) | Aug 2023 – Nov 2023
- Innovated a healthcare data analytics platform by integrating data sources, optimizing backend workflows, and implementing efficient ML algorithms.
- Improved data processing efficiency by 30% and improved disease detection accuracy to 85%+ by streamlining ETL processes and optimizing SQL queries.
- Collaborated with front-end developers to design a user-friendly interface using JavaScript, integrating custom DL models for disease detection, resulting in a 20% increase in user engagement and enhanced platform versatility, employing Agile methodology.
- Engineered a Socratic AI teaching assistant using FastAPI, LangChain, and Groq’s Mixtral-8x7B, processing 1000+ student queries with dynamic, question-driven learning. Integrated Google Search, Wikipedia, and YouTube APIs to enhance real-time knowledge retrieval. Developed an interactive code editor for hands-on coding exercises with real-time execution.
- Built a multimodal AI system supporting text, image, video, and voice-based queries, leveraging fine-tuned LLMs for reasoning and response generation. Created real-time DSA visualizations, improving conceptual understanding for 500+ students, and implemented custom speech synthesis and OCR-based image analysis, optimizing multimodal interactions for 80%+ accuracy in educational queries.
- Improved student engagement by 60% through an AI-driven learning platform, delivering personalized, context-aware tutoring. Enabled scalable knowledge access, automating 90% of manual query resolution. The code editor and DSA visualizations reduced debugging time by 40%, enhancing learning efficiency for students worldwide.
- Designed and implemented a Flask-based backend to train neural networks on Boolean functions, integrating TensorFlow/Keras for model creation, training, and inference.
- Built a React-based frontend that dynamically accepts Boolean functions and neural network configurations, sending training requests to the Flask API and rendering a 3D scatter plot using PCA for dimensionality reduction.
- Optimized model training and deployment on Render (0.1 CPU), supporting multiple activation functions (ReLU, Sigmoid, Tanh, Softmax), loss functions (MSE, Binary/Categorical Cross-Entropy), and optimizers (Adam, SGD) while ensuring low computational overhead.
- Implemented a secure online exam platform with real-time proctoring using FaceDB and OpenCV to detect and flag suspicious activities.
- Built an efficient backend system using Flask and FaceDB, ensuring seamless user authentication and secure data storage.
- Key Metrics: Face Recognition Module achieved ~99% TPR, 0.1%-1% FPR, ~98%-99% Accuracy, ~95%-99% Precision, ~97%-99% F1 Score, and an AUC of >0.99. Default Euclidean threshold set to 0.6 for recognition.
- Impact: Improved exam integrity by 40% and offered an efficient solution for remote education institutions to conduct exams securely.
- Designed a blockchain-based Merkle Proof Verifier smart contract to validate transactions with a cryptographic hashing mechanism, reducing verification time by 30% compared to naive approaches.
- Developed the contract in Solidity, tested integration with Web3.js, and deployed on a local Ethereum network to handle up to 10,000 transactions efficiently with minimal gas costs.
- Impact: Achieved scalability improvements of 20%, enabling secure and cost-effective transaction verification for decentralized systems.
- Innovated an end-to-end healthcare platform for detecting Alzheimer's stages from MRI scans, incorporating a login and complaint registration system. Integrated ML models like Custom CNNs, VGG16, and ResNet50 to optimize disease detection.
- Designed a user-friendly interface using HTML3, CSS, JavaScript, and Bootstrap, and optimized backend systems using Flask and MySQL. Streamlined ETL processes and improved data processing efficiency by 30%.
- Impact: Achieved 85%+ disease detection accuracy, increased user engagement by 20% through a seamless interface, and enhanced platform versatility for real-world healthcare applications.
- Built a machine learning model using a Gaussian Classifier to classify individuals into houses from the Harry Potter Universe based on the Big Five Personality Score.
- Achieved an 81.34% model accuracy, showcasing the effectiveness of personality-based predictive analysis.
- Impact: Engaged users with an interactive personality classification system while demonstrating real-world ML applications in entertainment.
- Developed a genome sequence classification model using Multinomial Naive Bayes and NLP to classify genomes into classes like chimpanzees, humans, and dogs.
- Impact: Delivered a highly accurate solution with classification rates of 99.8% for chimpanzees, 98.6% for humans, and 92% for dogs, emphasizing its capability for genetic research and analysis.
You can reach out to me via
