Professional Experience

  1. Home
  2. Industry

Experience Summary

Machine Learning, Signal Processing, and Algorithms
  • Low-Level Sensor Signal Processing: Engineered real-time signal processing pipelines to filter and extract high-fidelity features from raw sensor data, significantly optimizing downstream touch input detection.
  • Classical ML & Production Lifecycle: Curation of specialized datasets, feature engineering, and deployment of classical ML models (e.g., XGBoost, Random Forests) for real-time finger detection and classification, implementing active monitoring for model drift.
  • Deep Learning for Touch & Spatial Calibration: Architected CNN models to automate complex sensor calibration and high-accuracy palm recognition.
  • Multi-Touch Force Analytics: Developed ML algorithms and spatial interpretation models to resolve tactile force inputs, achieving precise per-finger force distribution across active touch surfaces.
  • Real-Time Cognitive Load Analytics: Developed deep learning models for real-time cognitive load estimation from eye-tracking data, featuring custom gaze preprocessing, feature extraction, and low-latency inference.
  • Class Imbalance & Small-Data Generalization: Implemented advanced data augmentation and loss-weighting strategies (e.g., SMOTE, focal loss) to address severe class imbalance in perception models, improving cross-user generalization under sparse data conditions.
  • Personalization & Low-Latency Model Optimization: Evaluated and deployed real-time warm-start fine-tuning for adaptive personalization across cognitive load and drowsiness models, while also implementing quantization techniques to achieve low-latency, edge-side inference.
  • Time-Series Sequence Modeling: Implemented recurrent and sequence models (LSTM/GRU/Transformers) to infer temporal fatigue dynamics from high-frequency eye-tracking time-series data.
  • Robust Multi-Tier Classification Architecture: Formulated a multi-tier classification framework that heightened perception model accuracy and stability across heterogeneous task environments and user demographics.
Computer Vision
  • Vision Transformers for Gaze & Facial Analytics: Designed custom Vision Transformer (ViT) architectures for real-time facial gesture recognition and gaze estimation, implementing novel attention mechanism adaptations and low-latency optimization strategies.
  • Clinical Ocular Diagnostics: Architected and fine-tuned a Vision Transformer (ViT) model to extract fine-grained spatial features from eye imagery, achieving high-accuracy automated detection and classification of clinical eye conditions.
  • Real-Time Touchpad Gesture Modeling: Engineered and optimized lightweight CNN architectures for continuous finger tracking and multi-finger gesture recognition, enabling real-time, dynamic performance adaptation.
  • Semantic Segmentation for Spatial Touch Control: Implemented a U-Net–based semantic segmentation model to accurately mask resting palm regions on touch surfaces, driving real-time adaptation of click-force sensitivity.
  • Industrial Vision & Inspection Systems: Implemented classical image processing and shape recognition algorithms deployed directly to industrial automated inspection systems.
  • OCR & Document Intelligence: Built an end-to-end Optical Character Recognition (OCR) pipeline incorporating custom image segmentation algorithms and multi-class classifier architectures.
Human-Computer Interaction, Eye Tracking, and Wearable Computing
  • Eye-Tracking & Gaze-Interaction Systems: Spearheaded R&D in eye tracking and spatial gaze analytics for real-time human-computer interaction and predictive user behavior models.
  • Multimodal Input Frameworks: Architected a novel multimodal interaction paradigm combining real-time gaze tracking and foot-based inputs to enable fluid point-and-click interactions and dwell-free typing.
  • Biometric Authentication & Gaze Security: Engineered a robust gaze-gesture authentication framework designed to counter shoulder-surfing and video-analysis threats, creating a unified repository of eye-gestures for cross-application interaction.
  • Immersive VR & Accessibility Interaction: Researched and deployed gaze-assisted interaction models in virtual reality environments, evaluating the impact of dynamic interface layout, selection mechanics, and field-of-view motion on typing throughput and usability.

Technology, Tools, and Frameworks

  • Programming Languages & Runtimes: Python, C++, C#, Java, MATLAB, SQL, CUDA, Node.js, Qt C++.
  • Deep Learning: Deep Neural Networks, CNN, TCN, GRU, LSTM, RNN, CNN-GRU Hybrids, U-Net, Autoencoders, Transformers, Attention Mechanisms, Encoder and Decoder, Time Series Modeling, Sequence Modeling, Multi-task Learning, Transfer Learning, Domain Adaptation, PyTorch, CUDA, ONNX.
  • Classical Machine Learning: XGBoost, Gradient Boosting, Random Forests, Supervised Learning, Unsupervised Learning, Regression, Classification, Clustering, Feature Engineering, Dimensionality Reduction (PCA, t-SNE, UMAP), Regularization, Cross-Validation, Hyperparameter Optimization, Class Imbalance Handling (SMOTE, Class Weighting), Scikit-learn, NumPy, Pandas, SciPy, Seaborn, Matplotlib.
  • Computer Vision: CNN-based Computer Vision Models for object detection and localization, Vision Transformer (ViT), Vision-Language Models (VLM), R-CNN, YOLO models, MediaPipe, Pose Estimation, Hand Gesture Recognition, OpenCV for Image Segmentation, Object Detection, Tracking, Counting, Optical Character Recognition, Shape Detection.
  • Signal Processing and Algorithms: EEG, EOG, Eye Tracking, Pupillometry, Filtering, Adaptive Thresholding, Resampling, Feature Extraction, Sketch Recognition, Template Matching, Gesture Recognition, Activity Recognition.
  • User Research: Expertise in designing and conducting user studies for system evaluation (quantitative and qualitative), performing statistical analysis, and hypotheses testing through parametric and non-parametric methods.

Sensel, R&D | Sunnyvale, CA, USA

Touch Tracking, Force Sensing, and Haptics Solutions

Languages, Tools & Environment: C, C++, Python, C# .NET, Visual Studio, Windows, Linux
Responsible for conducting research and development toward inventing solutions for human interface design to create and improve the world’s most advanced touch technology. Research focused on building core competencies of Sensel in the areas of accurate force sensing, precise touch tracking, and creating rich haptic experiences. Worked cross-functionally with engineering and design teams to integrate these novel technologies into consumer electronic devices such as laptop touchpads, automotive interactive surfaces, gaming peripherals, virtual reality, and more.
Key Roles & Responsibilities
  • Develop new algorithms to enhance touch and sensor performance across a variety of hardware and use-cases.
  • Apply image processing, machine learning and related techniques to interpret low-level sensor data.
  • Develop and evaluate new user experience concepts and prototypes that showcase Sensel’s force-sensing touch technology.
  • Design and run user studies to set target accuracy and sensitivity metrics.
  • Operate at all levels of software stack, from firmware to end-user applications.
  • Build diagnostic tools and test infrastructure to streamline sensor development.
  • Develop automated software and robotic sensor tests to validate quality of algorithm output.
  • Customize software builds to support new devices, form-factors, and prototypes.

National Instruments, R&D | Bangalore, India

Machine Vision Development Module

Languages, Tools & Environment: C, C++, C# .NET, LabVIEW, Visual Studio, Windows, Linux, VxWorks
Development of image processing algorithms for "Machine Vision" systems. Further, it involved Platform Portability and Multi-coring of Vision Algorithms on various real-time architectures.
Key Roles & Responsibilities
  • Pursue research and development in distinctive machine vision application areas.
  • Platform portability and Multi-coring of Vision Algorithms.
  • Developing API interfaces for various programming frameworks (technologies) to leverage the functionalities of vision development module.

Computer Vision - Optical Character Recognition (OCR)

Languages, Tools & Environment: C, C++, C# .NET, LabVIEW, Visual Studio, Windows, Linux, VxWorks
Development of algorithms for "Optical Character Recognition" (OCR) system for robotics and industrial applications. The OCR module implements intelligent algorithms for Thresholding, Segmentation, and Classification to enable optical character recognition.
Key Roles & Responsibilities
  • Developed algorithms for character Segmentation based on various heuristics.
  • Implemented Shortest Path Graph Cut algorithm to make character Segmentation process smart enough to resolve complicated cases involving merged characters, low-intensity images, broken characters, and multiline detection.

NOKIA R&D | Bangalore, India

FlashLite 4.x Plugin for Qt Browser on Qt WebKit

Languages, Tools & Environment: Qt Framework, Qt C++, Symbian C++, Carbide.c++, SVN, Rational Synergy, Rational Change, Symbian^3
Development of Qt FlashLite Plugin for Nokia Qt Browser on Qt WebKit. The plugin enabled high-performance video playback from Web-video platforms like YouTube and Google Videos.
Key Roles & Responsibilities
  • Research and development towards implementing Adobe Flash Player on Symbian platform.
  • Worked closely with Adobe to implement the Qt Flash plugin for mobile browsers, taking it from idea to prototype and full production.
  • Worked extensively on the development of the Gesture Recognition module and implemented an algorithm to improve web page scrolling performance by exchanging bitmaps with the browser rendering engine and dynamically switching between software and hardware rendering modes.

Implementation of FlashLite 4.x Plugin and Flash Player on Symbian^3

Languages, Tools & Environment: Symbian C++, Carbide.C++, SVN, Rational Synergy
Development of the FlashLite 4 plugin and standalone Flash Player on the Symbian^3 platform, enabling standalone playback of Flash video and audio files as well as web video streaming.
Key Roles & Responsibilities
  • Research and development towards implementing Adobe Flash Player on Symbian^3 platform.
  • Worked on the development of a standalone flash player with hardware rendering capabilities.

Robert Bosch Engineering & Business Solutions | Bangalore, India

Measure Data Analyzer (MDA)

Languages, Tools & Environment: VC++, Borland C++, VCL, COM Programming on Windows 2000/XP/Vista
Measure Data Analyzer (MDA) software is an offline calibration and signal analysis tool for displaying and analyzing saved measurement data of various automotive engine parameters of a vehicle.
Key Roles & Responsibilities
  • Implemented different ways to visualize high-resolution measurement data and developed algorithms to interpolate and extrapolate missing data.
  • Developed an intuitive print layout editor to enable calibration engineers to share and analyze signal data, generating vectorized image outputs that maintain clarity at high zoom levels.

NDS Private India Limited (Acquired by CISCO) | Bangalore, India

Game Studio

Languages, Tools & Environment: C# .NET programming on Windows 2000/XP
Game Studio is a Set-Top Box application that generates a virtual 3D canvas for games running on set-top boxes.
Key Roles & Responsibilities
  • Defining use cases, architecture design, core implementation, and test suite creation.