Machine Learning Software Engineer | GenAI Developer | AI Researcher
I'm a Machine Learning Software Engineer with 4+ years of experience delivering production-grade ML systems, GenAI applications, and cloud-native solutions. I specialize in GenAI (LLMs, RAG, LangChain), deep learning (PyTorch, TensorFlow), robotics & BCI (Kinova Gen3, EEG motor imagery), and full-stack development (React, Next.js, Node.js).
My work spans building CNN-based robotic control systems achieving 70% accuracy and co-authoring peer-reviewed research in IEEE Access 2025 on multimodal emotion detection (98% accuracy). I leverage AWS (SageMaker, Bedrock, Lambda), MLOps (Docker, CI/CD), and big data tools (PySpark, Hadoop) to design scalable, real-time AI solutions that bridge research and real-world impact.
Master of Science in Data Science | University of Maryland, Baltimore County (UMBC)
Jan 2024 – Dec 2025 | GPA: 3.75/4.00
Coursework: Machine Learning, Deep Learning, NLP, Data Engineering
Bachelor of Engineering in Computer Engineering | Gujarat Technological University
Jun 2019 – May 2023 | GPA: 3.5/4.00
Coursework: Data Structures & Algorithms, Database Systems, OOP, Web Development
Deep Fusion of Neurophysiological and Facial Features for Enhanced Emotion Detection
Safavi F., Parikh D. et al. | IEEE Access, vol. 13, pp. 67434–67445, 2025
DOI: 10.1109/ACCESS.2025.3555934
Developed multimodal transformer integrating EEG and facial data, achieving near state-of-the-art accuracy.
CNN-based EEG classification for Kinova Gen3 robotic arm control with 70% accuracy. Built PyTorch (CUDA) models with REST API integration, simulated in Isaac Sim for real-time control pipelines.
Deep-RNN model detecting emotions from DEAP EEG signals with ~62% accuracy. Applications in human-computer interaction and affective computing.
🔗 GitHub
CNN model for histopathology image classification using transfer learning. Demonstrates AI-assisted medical diagnosis potential.
🔗 GitHub
Face detection system using OpenCV Haar Cascades with live webcam feed processing. Optimized for low-latency detection.
🔗 GitHub
ML model predicting transfer values using ensemble and neural networks. Reduced forecast error by 37.4% with 99.95% accuracy.
🔗 GitHub
Big data risk profiling for unmanned aerial systems using PySpark, AWS Glue, and Hadoop. Analyzed 100+ GB of FAA data with 95% accuracy.
🔗 GitHub
- Software Engineer Certification - HackerRank
- Building Real-Time Video AI Applications - NVIDIA
- AWS Educate Machine Learning Foundations - AWS