Featured Projects
Autonomous Drone Navigation & Hardware Deployment
Constructed 3 and 5-inch quadcopters with custom 3D-printed frames for physical hardware deployment. Integrated an onboard Raspberry Pi to run pre-trained PPO reinforcement learning models for real-time autonomous flight navigation within an OptiTrack environment.
Power-Aware Co-Optimization of Unconventional Hexacopters via Evolution and Learning
Designed a bi-objective evolutionary framework using NSGA-II to co-optimize physical hexacopter morphologies and neural network controllers trained via PPO. Modeled realistic LiPo battery dynamics and executed large-scale parallel simulations on a computing cluster, achieving a 31.6% energy reduction while preserving 95.7% of flight performance.
Pulse - personalized content feed
Architected a full-stack personalized content feed using Flask and SQLite, aggregating data from GNews, YouTube, and Lemmy APIs. Developed an NLP-driven ranking engine with sentiment analysis and cross-source clustering to group multi-perspective stories, backed by a robust suite of over 325 tests.
Autonomous Solar-Powered Mars Rover
Designed and synthesized a complete digital control system in VHDL on an Artix-7 FPGA for autonomous robotic navigation. Engineered a custom physical power management system featuring a buck converter, supercapacitor buffer, and a solar panel array.
Voice-Driven Conversational AI Pipeline
Engineered an end-to-end voice-interactive recipe recommendation system hosted on a custom web interface. Fine-tuned a BERT transformer model for intent classification and integrated it with a Prolog knowledge base, MARBEL dialogue manager, and OpenAI Whisper for seamless speech processing.
Deep Learning Architecture
Constructed a LeNet-5 CNN from scratch using pure NumPy with Numba JIT compilation for a 12x execution speedup. Implemented a custom ResNet-18 model in PyTorch and benchmarked multiple architectures on the MNIST dataset using bootstrap statistical methods.
Predictive Health Analytics & Stress Modeling
Processed and merged multi-source smartwatch and lifestyle datasets to analyze the impact of sleep stages and habits on overall efficiency. Trained and evaluated machine learning classifiers like Decision Trees to predict daily stress categories based on heart rate variability and sleep data.