Resume

Education

Doctoral Candidate, Chair of Information Oriented Control, TUM, Oct 2020 - Present

• Supervised by Prof. Sandra Hirche
• Thesis: Data-Driven Analysis of Human Motor Behavior for Medical Applications

  MSc. Informatics, Technical University of Munich, Aug 2019

• Thesis (Conducted at {Volkswagen Group AI Research, Germany}): Learning State-Space Models of Camera-Based Robots for Intrinsically Motivated Control
• Passed with High Distinction.

  BTech. Software Engineering, Delhi Technological University, May 2013

• Final Project: Tower Defense Game Implementing Bee Colony Algorithm.
• First Division with Distinction.

Work experience

Research Associate, Technical University of Munich, Germany, Oct 2020 - Present

• Contributions to Horizon 2020 project ReHyb:
  • Developed explainable, data-driven algorithms for detecting anomalous motion in stroke patients using unsupervised anomaly detection techniques and generative modeling.
  • Designed inverse optimal control methods to mitigate compensatory (abnormal motion by stroke patients) movements via robotic feedback.
  • Implemented a data-driven framework for enabling FES-assisted motion through automatic detection of intended movements from muscular activity.
  • Led experimental studies - protocol design, participant recruitment, and creation of web-based tools for motion data collection and labeling. Acquisition and processing of EMG-based, optical-marker and video-based datasets from healthy and post-stroke participants for model training and validation.
• Contributions to Horizon 2020 project ConPDMode:
  • Implemented and compared machine learning models (e.g., Gaussian Processes, ensemble, and deep networks) for Parkinson’s Disease (PD) symptom severity classification from wearable IMU data and Identified key challenges in their application, including uncertainty estimation and data imbalance.
  • Proposed a novel uncertainty quantification approach for deep learning to detect data gaps via epistemic uncertainty estimation.
  • Designed methods to address dataset imbalance in PD symptom classification.
  • Designed the prototype for a web and mobile application for motion data visualization and PD symptom severity classification with interfaces for physicians and patients.
• Contributions to Horizon 2020 project CO-MAN:
  • Developed a safe, user-preference–driven navigation framework using Preferential Bayesian Optimization.
  • Contributed to designing a safe control approach for unknown dynamics with Control Barrier Functions.
• Supervision experience:
  • Supervised over 15 undergraduate and graduate student thesis, as well as additional research and engineering practice projects
  • assisted with course organization, teaching, exam design and evaluation for several courses.

AI Resident, Meta, U.S.A, Sep 2019 - Oct 2020

• Representation learning for robot manipulation: Contributed to the design of an extended body schema of a robotic arm to enable manipulation of held tools from visual and proprioceptive inputs.
• Model-Based Inverse Reinforcement Learning: Developed an inverse reinforcement learning approach inspired by meta-learning, using gradient updates to efficiently robot behaviors from visual demos by humans.
• Learning state-dependent losses for inverse-dynamics learning: Demonstrated that meta-learning adaptive loss functions improves inverse-dynamics model learning for a robotic arm compared to conventional fixed-loss approaches.

Software Developer, Epic Systems, U.S.A, Oct 2014 - Sep 2016

• Designed and implemented web-based interfaces for, as well as enhanced backend features of an EHR application supporting medical information summarization and diagnosis assistance.

Software Developer in Test, Intel Security, India, Jul 2013 - Sep 2014

• Created a framework in C++ for stress testing a whitelisting product for Windows systems.
• White-box tested the whitelisting product.

Skills

• Languages: Python, MATLAB, Simulink, C++, JavaScript, CSS, SQL
• Libraries: Pytorch, Tensorflow, Scikit-learn, OpenCV, Pandas, Matplotlib, Blender
• Tools: OpenSim, Blender, Visdom, Figma
• Concepts: Machine Learning, Explainable AI, Inverse Optimal Control, Reinforcement Learning, Generative AI, Anomaly Detection, Probabilistic Graphical Models, Learning from Feedback, Human-centered AI, Large Language Models

Selected Publications

A more extensive list can be found in Google Scholar.

Explainable Unsupervised Anomaly Detection for Identifying Compensatory Motor Behavior in Stroke Rehabilitation

N. Das, S. Endo, S. Patel, M. Rossini, A. De Crignis, E. Guanziroli, G. Palumbo, A. Specchia, C. Krewer and S. Hirche, "Explainable Unsupervised Anomaly Detection for Identifying Compensatory Motor Behavior in Stroke Rehabilitation," 2024 10th IEEE RAS/EMBS International Conference for Biomedical Robotics and Biomechatronics (BioRob), Heidelberg, Germany, 2024, pp. 1320-1327, doi: 10.1109/BioRob60516.2024.10719910

Framework for Learning a Hand Intent Recognition Model from sEMG for FES-Based Control

N. Das, S. Endo, H. Kavianirad and S. Hirche, "Framework for Learning a Hand Intent Recognition Model from sEMG for FES-Based control," 2024 10th IEEE RAS/EMBS International Conference for Biomedical Robotics and Biomechatronics (BioRob), Heidelberg, Germany, 2024, pp. 1320-1327, doi: 10.1109/BioRob60516.2024.10719910

Online detection of compensatory strategies in human movement with supervised classification: a pilot study.

Das, N., Endo, S., Patel, S., Krewer, C., & Hirche, S. Online detection of compensatory strategies in human movement with supervised classification: a pilot study." 2023 Frontiers in Neurorobotics, 17, 1155826

Deep Learning based Uncertainty Decomposition for Real-time Control.

Das, N., Umlauft, J., Lederer, A., Capone, A., Beckers, T., & Hirche, S. Deep Learning based Uncertainty Decomposition for Real-time Control." 2023 IFAC-PapersOnLine, 56(2), 847-853

Learning Extended Body Schemas from Visual Keypoints for Object Manipulation.

S. Bechtle, N. Das and F. Meier, "Learning Extended Body Schemas from Visual Keypoints for Object Manipulation." 2020 arXiv preprint arXiv:2011.03882.

Model-Based Inverse Reinforcement Learning from Visual Demonstrations

N. Das, S. Bechtle, T. Davchev, D. Jayaraman, A. Rai and F. Meier, "Model-Based Inverse Reinforcement Learning from Visual Demonstrations," 2020 Conference on Robot Learning (CoRL)

Learning State-Dependent Losses for Inverse Dynamics Learning

K. Morse, N. Das, Y. Lin, A. S. Wang, A. Rai and F. Meier, "Model-Based Inverse Reinforcement Learning from Visual Demonstrations," 2020 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS)

Beta DVBF: Learning State-Space Models for Control from High Dimensional Observations.

N. Das, M. Karl, P. Becker-Ehmck and P. van der Smagt, "Beta DVBF: Learning State-Space Models for Control from High Dimensional Observations." 2019 arXiv preprint arXiv:1911.00756.

3D Human Body Segmentation

Das, Neha. (2018). Development of a system that allows for the semantic segmentation of a 3D model of a human body into its constituent parts.

Seminar: Deep Learning Sequence Modelling (Natural Language Processing)

Das, Neha. (2018). Seminar: Deep Learning Sequence Modelling (Natural Language Processing).

Teaching