I'm Ganesha, a USC Trojan based in Los Angeles. I build ML systems that actually work in production — with a focus on edge deployment, optimization, and making AI practical and reliable.
Building production-ready models for edge deployment. Fine-tuned MobileNetV2 and MobileNetV4 with quantization, pruning, and staged transfer learning. Re-engineered MobileNetV4 for Akida-compatible conversion, analyzing power, memory, and latency trade-offs.
Co-Director of AI & Innovation Team at GRIDS Club. Coursework: Deep Learning, NLP, Machine Learning, Information Retrieval, Foundations of AI, Database Systems.
Developed emotion recognition models using TCNs, OpenFace, and MediaPipe. Deployed FastAPI services on AWS for scalable, real-time behavioral assessments.
Explored reframing clustering problems as SAT formulations. Applied DPLL algorithm across 2-SAT and 3-SAT, connecting theoretical CS to ML sparsity concepts.
Built a strong CS foundation and discovered my passion for machine learning.
AI teaching assistant that makes dense ML research papers easier to understand. Uses FAISS for semantic search across 200+ papers, parses PDFs with GROBID, and teaches section-by-section through a grounded RAG pipeline.
Multi-agent GenAI travel planner for accessibility-first travel. Helps travelers with wheelchair access, dietary needs, and budget caps get verified, realistic itineraries. Four specialized LLM agents handle constraints, venues, routes, and generation.
Transforms complex documents into interactive semantic mind maps with full source traceability. Instead of linear summaries, it creates navigable graph networks where concepts and relationships are visually mapped. Every node links back to exact evidence from the original document, eliminating hallucinations.
Always open to new opportunities, collaborations, or just a good conversation about ML and tech.