Developing an ultrasound-based directional audio system using DSP for audio modulation and a Class D op-amp for amplification. Migrating legacy DSP firmware to the latest Code Composer Studio (CCS) for development and debugging. Designing PCB, integrating DSP and Class D amplifier for improved performance and miniaturization.

Conducting an extensive literature review on multi-scale CNN architectures, analysing advancements, applications, and challenges. Synthesizing insights to identify research gaps and propose future directions in multi-scale deep learning models.

Designed and implemented a data-driven questionnaire to optimize user profiling and enhance match accuracy. Developed backend systems for algorithmic matching using LLM, based on user preferences for 150+ users.

Built models (Decision Tree, Random Forest, XGBoost, etc.) to predict earthquake severity with 98.5% accuracy. Conducted data cleaning, feature engineering, and model evaluation for disaster prediction.

Engineered and integrated a closed loop audio amplifier system (THAT2180C VCA, LM380N PA, CA3140 VU Meter) on an STM32+VScope platform calibrating subsystem gains/offsets and characterizing PA frequency response (100Hz to 50kHz) while developing MicroPython/Python (Jupyter Notebook, ipywidgets) firmware for real time RMS feedback and step up/down volume stabilization

Built a Streamlit interface for PDF uploads that splits texts into overlapping chunks and issues HTTP POSTs to a local Ollama endpoint for LLM inference. Used Chroma to generate and index embeddings—wrapped in an LRU cache (256 entries) to skip redundant encoding calls. Performed cosine-similarity searches to fetch the top-5 most relevant passages from Bhagavad Gita commentaries with source attributions.