Rushikesh Kusuma

Rushikesh Kusuma- |-| linkedin.com/in/rushikesh | github.com/rushi-k12 Education ABV-IIITM Gwalior Nov. 2022 – Present B.Tech in Information Technology FIITJEE Junior College 7.96 CGPA Apr. 2020 – June 2022 Higher Secondary 97.2% Experience IIT Mandi iHub & HCI Foundation May 2025 – Present Research Intern IIT Mandi, India • Contributing to Wellness Sense, an AI-powered lifestyle management system (Ref: IIT MANDI/iHub/RD/2024-25/02). • Prototyping IoT-based Human Activity Recognition system using multi-modal sensor fusion & edge analytics. • Building data acquisition pipeline and selecting hardware for real-time inference and predictive modeling. • Working on wearable sensor data integration with behavioral analytics for personalized health insights. • Focused on system architecture, hardware procurement, and early-phase testing. Coding Pro June 2024 – Aug 2024 AI/ML Developer Intern Remote, India • Developed chatbot using advanced NLP for intelligent text-based interactions and coding assistance. • Implemented RetrievalQA with Mixtral-8x7B for precise question-answering, improving user satisfaction. • Integrated Ollama models for local inference in VS Code extension, enabling real-time autocompletion and code suggestions. • Integrated Groq API to enhance chatbot performance with high-performance cloud-based inference. • Built Streamlit interface for seamless user interaction with extension features and chatbot. Projects Bitemporal 3D Change Detection using LiDAR and Machine Learning May 2025 – July 2025 • Tech Stack: Python, Open3D, NumPy, scikit-learn, XGBoost, OpenCV, Matplotlib, Raspberry Pi 4B • Developed a complete pipeline for detecting semantic and geometric changes between two 3D LiDAR scans (epoch0 epoch1) using statistical outlier removal, voxel downsampling, and ICP alignment. • Engineered spatial and color-based features (pointwise distance, neighborhood distance, RGB change) to train models for change classification using Random Forest, XGBoost, and KNN. • Benchmarked models on accuracy, inference time, memory usage, and deployed optimized model on Raspberry Pi 4B for real-time point cloud change detection. LifeScan: Noise-Free Imaging and Body Part Recognition Jan. 2025 – Apr. 2025 • Tech Stack: Python, ONNX, TensorFlow, Raspberry Pi, Matplotlib, PIL • Designed deep learning pipeline with autoencoder for image denoising and CNN for body part classification. • Converted models to ONNX, deployed using onnxruntime, achieving 92% test accuracy and 0.0204 validation MSE. • Benchmarked and deployed models on Raspberry Pi 4B, optimized for low-latency, low-memory environments. Technical Skills Languages: Java, Python, JavaScript, C/C++, SQL Web/Frameworks: Node.js, Express.js, React.js, Redux, MongoDB, JWT, REST APIs Libraries/ML Tools: TensorFlow, ONNX, NumPy, Pandas, Matplotlib, PIL Developer Tools: Git, VS Code, PyCharm, IntelliJ, Render, MongoDB Atlas, Cloudinary, PayPal SDK Cloud Platforms: AWS, Google Cloud Platform Visualization: Power BI, MS Excel Achievements • • Edge Innovation AI Challenge 2024 Oct 2024 Top 10 Finalist — Autonomous Fire Extinguishing Vehicle, Top 10/200 teams, DigiToad and STMicroelectronics. Microsoft Certification July 2024 Earned Microsoft Certified Azure Data Fundamentals (DP-900) credential.