cv
Summary
| Overview | PhD-level Machine Learning Scientist with 5+ years of experience developing end-to-end deep learning frameworks for complex time-series and physiological signal analysis. Proven ability to architect novel CNN and Transformer models that deliver real-time, interpretable, and scalable solutions from raw sensor data. Passionate about translating multimodal biosignals into discernable insights for MedTech and human performance applications. |
Technical Skills
| Languages & Frameworks | Python (TensorFlow, PyTorch, JAX, Scikit-learn, MNE), MATLAB, SQL, R, Git |
| ML Architectures | Transformers, CNNs (1D-CNN, WaveNet), RNNs/LSTMs, Autoencoders, Foundation Models |
| ML Concepts | End-to-End Learning, Self-Supervised Learning, Time-Series Analysis, Explainable AI (XAI), Attention Mechanisms, Few-Shot Learning, Sensor Fusion |
| Domain Knowledge | Biosignal Processing (fNIRS, EEG, ECG), Brain-Computer Interfaces (BCI), Clinical Study Design, Human Motor Skill Assessment |
Research Experience
-
Sep 2025 – Present
Postdoctoral Researcher
CeMSIM, RPI
- Pioneering multimodal learning by integrating auxiliary physiological signals (HRV, pupillometry) with neuroimaging data to build a comprehensive model of human performance.
- Expanding the model's core function from binary classification to regression, enabling the prediction of precise, quantitative scores for surgical certification (FLS).
- Developing advanced modeling techniques for high-bitrate time-series data, focusing on adapting Transformer architectures for complex EEG signal analysis.
-
Jan 2020 – Aug 2025
Graduate Researcher, AI/ML (Health)
CeMSIM, RPI
- **Architected and validated a novel 1D-CNN framework** that achieved **98.6% accuracy** in classifying motor skills from raw fNIRS data, enabling real-time analysis.
- **Designed and implemented a Transformer-based foundation model** with novel attention mechanisms, providing spatiotemporal explanations that transformed a "black box" model into an interpretable diagnostic tool.
- **Demonstrated state-of-the-art generalization**, adapting the foundation model to a novel medical task with **>87% accuracy using fewer than 30 labeled samples**, proving the model's scalability.
- **Owned the end-to-end creation of a foundational 2,100+ trial neuroimaging dataset**, managing curation and processing of 100GB+ of data to enable training a first-of-its-kind foundation model.
Professional Experience
-
May 2018 – Sep 2023
R&D Co-Director & Co-Founder
GRIT Engineering Pvt. Ltd.
- Co-founded an engineering firm, leading a cross-functional team of 5+ engineers and technicians from concept to delivery of custom electromechanical systems.
- **Directed the full project lifecycle** for a high-visibility automated camera system for "The Voice Nepal," managing client requirements, system integration, and successful on-time deployment.
- Owned the R&D roadmap, driving iterative product improvements through rigorous design testing and root cause analysis, directly impacting system reliability and client satisfaction.
Education
-
Aug 2025
Ph.D. in Mechanical Engineering
Rensselaer Polytechnic Institute
- Dissertation: "*End-to-end bimanual motor skill assessment from raw neuroimaging data*"
-
May 2024
M.Eng. in Mechanical Engineering
Rensselaer Polytechnic Institute
Select Publications
- **A. Subedi** et al. “An Interpretable Transformer-Based Foundation Model for Cross-Procedural Skill Assessment...”, *arXiv:2506.22476*, 2025.
- **A. Subedi** et al. “End-to-End Deep Learning for Real-Time Neuroimaging-Based Assessment of Bimanual Motor Skills”, *npj Digital Medicine (Under Review)*, *arXiv:2504.03681*, 2025.
- C. Eastmond, **A. Subedi**, S. De, X. Intes, “Deep Learning in fNIRS: A Review”, *Neurophotonics 9(4).*
Professional Activities & Awards
- **Peer Reviewer** for scientific journals including *Neurophotonics* and *Heliyon*.
- **Teaching Assistant** for Numerical Methods & Machine Dynamics (RPI, 2020).
- ABU Robocon (Asia-Pacific Robot Contest): Team Nepal (**Best Engineering Awards: 2015, 2016**).
- IOE Entrance Scholarship: Full-tuition, merit-based award.