Pramil Paudel
Research Engineer · PhD Candidate privacy-preserving vision / lensless imaging / ML systems

Research-minded engineer. Production-minded researcher.

I am a research engineer and PhD candidate working at the intersection of machine learning, computational imaging, and privacy. My research focuses on building learning systems that operate directly on encoded sensor measurements rather than human-viewable images.


Before my PhD, I built production systems from the ground up, as a software engineer, team lead, engineering manager, and backend solutions architect. I have led teams, designed backend and data platforms, built cloud pipelines, and scaled systems across healthcare, insurance, analytics, and research environments. That is the part I bring into research: I like hard problems, but I also like building the complete system around them.

View Research Systems Research Outputs CV Email
5+ Years Industry Solutions Architect and Engineering Lead — distributed systems, cloud infrastructure, and 50+ engineers trained across multiple companies.
PhD Research Privacy-preserving computer vision and lensless imaging — combining deep learning, adversarial ML, and computational imaging.
End-to-End Builder From embedded robotics to cloud microservices to deep learning pipelines — ships both research systems and production software.
Worked Domains High-Energy Physics · U.S. Healthcare · Auto and Commercial Insurance · Computer Vision · Deep Learning
Research Highlight

Seeing Without Revealing

From scene to encoded measurement to privacy analysis.

A research overview of the full lensless privacy pipeline.

1. Sensing: how ordinary scenes become lensless measurements.
2. Learning: classification or representation learning without direct reconstruction.
3. Communicating: lensless representation embedding inside natural RGB containers.
4. Attacking: reconstruction attempts and privacy leakage evaluation.
Research Systems

Research systems.

Reconstruction-free classification from lensless sensor measurements
System 01 · Semantic Understanding Without Seeing

Reconstruction-Free Classification for Lensless Imaging Systems

Learning semantic information directly from lensless sensor measurements without first reconstructing a human-viewable image.

QuestionCan a model understand visual content from direct sensor measurements that humans cannot interpret?
FocusReconstruction-free classification, sensor-domain learning, and frequency-aware representation analysis.
VenuesICMLA 2025
Lensless Imaging Sensor Measurements Reconstruction-Free Semantic Understanding
Lensless Hide
System 02 · Privacy-Preserving Diffraction Pattern Hiding

Privacy-Preserving Image-to-Image Steganography

Finding a way to communicate lensless images while preserving privacy and defending against reconstruction threats — hiding full diffraction representations inside natural RGB containers.

QuestionCan lensless measurements be transmitted covertly without exposing them to reconstruction attacks?
FocusImage-to-image steganography, diffraction pattern hiding, and reconstruction threat modeling.
Under Review Steganography Privacy Lensless Imaging
Lensless Reconstruction
System 03 · Reconstruction–Semantic Tradeoff Understanding

Privacy-Preserving Robust Lensless Image Classification

Finding the tradeoff between visual reconstruction risk and semantic understanding in lensless classification systems.

FocusHow much semantic utility remains when reconstruction becomes difficult?
VenueIEEE MASS 2024
Robust Classification Privacy Tradeoff Lensless Images
Physics-informed lensless diffraction understanding
System 04 · Physics-Informed Lensless Diffraction Understanding

Physics-Informed Lensless Image Diffraction Understanding

Interpreting lensless diffraction measurements using optical forward models, PSF structure, and physics-guided learning.

Question How can optical physics and learned models help explain what information is preserved inside lensless diffraction measurements?
Focus PSF structure, diffraction-domain analysis, optical forward modeling, and physics-guided interpretation of lensless measurements.
PSF Diffraction Physics-Informed ML Forward Models
Research Outputs

Papers · Book · Code

Robust Privacy-Preserving Classification for Lensless Images

IEEE MASS 2024

Sumit Bhattarai, Pramil Paudel, Zhu Li, Bo Luo, Fengjun Li

Peer-reviewedLensless ImagingPrivacy

Reconstruction-Free Lensless Image Classification

ICMLA 2025

Pramil Paudel, Fengjun Li

Peer-reviewedClassificationSensor Domain

Privacy-Preserving Image-to-Image Steganography

Under Review

Pramil Paudel, Bo Luo, Fengjun Li

PreprintSteganographyPrivacy

Realtime Anomaly Detection using Event-by-Event SVD

Preprint

Pramil Paudel, Daniel Tapia Takaki

PreprintHEPReal-time ML

Basic Electronics Engineering

Book

Bikal Adhikari, Pramil Paudel · Heritage Publication, Kathmandu

BookElectronics
Experience

Industry engineering plus graduate research.

May 2021–Present

Graduate Teaching Assistant & PhD Researcher

University of Kansas  ·  Advisor: Prof. Fengjun Li
  • RFLIC — Reconstruction-Free Lensless Image Classification. Improved lensless image classification without reconstruction using Block DCT, achieving state-of-the-art accuracy directly in the sensor measurement domain. [ICMLA 2025]
  • PPLiC — Robust Privacy-Preserving Classification for Lensless Images. Defeats GAN-based reconstruction attacks while maintaining semantic utility and preventing privacy leakage. [IEEE MASS 2024]
  • LiGANS — Lensless Diffraction Pattern Hiding via Steganography. Novel covert embedding of full lensless representations inside natural RGB containers without visible artifacts. [Under Review]
  • LiRT — Physics-Informed Lensless Image Reconstruction in the Wild using Fourier Neural Networks and optical forward models leveraging PSF structure. [In Progress]
  • Polarimeter — High-Energy Physics data processing platform that calibrates and estimates energy received by particle detectors within seconds of a spill. [T2 Lab · CERN]
  • HW-RankReduce — Hardware-Accelerated near real-time energy fluctuation detection using advanced rank reduction and event-by-event SVD at CERN scale. [Preprint]
PyTorch Lensless Imaging Privacy-Preserving ML GANs Fourier Neural Networks SVD Python
Mar 2019–Dec 2019

Backend Solutions Architect

Renegade Insurance, Kathmandu — formerly Covered by Sage, now a large-scale U.S. digital insurance platform
  • Designed and delivered the company's first production backend architecture from scratch: a real-time P&C insurance platform serving agents and customers across the United States.
  • Established the engineering culture, development standards, CI/CD pipelines, and system architecture that the company continues to build on today.
  • Built B2C and B2B integration layers connecting insurance agents, customers, and carriers for real-time policy quoting, underwriting, and support.
  • Delivered a fully functional, live platform within 12 months that remains actively serving clients nationwide.
  • Trained and mentored 50+ engineers in system design, microservices, cloud deployment, and Agile workflows, transforming the organization's engineering capability within a single year.
Spring Boot Spring Cloud Microservices Azure Kubernetes MongoDB MySQL Jenkins CI/CD
Mar 2017–Dec 2018

Engineering Manager

National Incubation & Research Center (NIRC), Kathmandu
  • Hired, trained, and managed a team of 20 to 25 engineers, QA specialists, and DevOps engineers, responsible for every aspect of engineering from coding standards to IT infrastructure.
  • Architected and built data analytics and visualization platforms using Groovy on Grails and homegrown Nepali datasets across multiple domains.
  • Designed AWS-based automation pipelines enabling scalable web deployment and continuous integration, bringing the organization to modern cloud practices.
  • Resolved complex cross-functional technical challenges across development, integration, and deployment; acted as the final technical authority across all projects.
  • Delivered multiple major products in short development cycles with high quality and on-schedule releases.
Java Groovy on Grails Python Hadoop Apache Spark AWS Jenkins Git
Dec 2014–May 2017

Software Engineer Team Lead

Deerwalk Inc., Kathmandu — largest software company in Nepal, U.S. healthcare data analytics
  • Joined as a software engineer and was promoted to Team Lead within one year through consistently exceptional performance.
  • Key products: Plan Analytics (PA) and Care Manager (CM), enterprise platforms delivering financial and clinical analytics on large-scale U.S. healthcare data.
  • As Lead Engineer on Plan Analytics, owned the full data ingestion pipeline: consumed thousands of raw medical data files from multiple vendors, enforced HIPAA-compliant workflows, and produced structured outputs for downstream front-end analytics teams.
  • Singlehandedly migrated a legacy MySQL-based data processing engine to Java and Hadoop, cutting end-to-end processing time from 30 days to 8 hours — a 90x performance improvement.
  • Designed Scala-based Apache Spark frameworks for large-scale healthcare dataset processing and custom client analytics reports.
  • Led teams in data scrubbing and transformation using the Cascading Framework for high-volume insurance claim data.
Java Scala Apache Hadoop Apache Spark Cascading MySQL AWS Python
Jan 2016–Dec 2016

Senior Software Engineer (Co-founder)

Yotcha Inc., Singapore / Kathmandu — Airbnb-style real estate platform for Singapore
  • One of 4 to 5 founding engineers who built the entire platform from scratch: a full-featured real estate listing engine covering all Singapore properties, with history tracking, search, and listing management.
  • Worked as a full-stack engineer across backend services, frontend UI, data pipelines, and deployment infrastructure.
  • Established development standards, CI/CD workflows, and automated deployment pipelines that defined the engineering culture from day one.
  • Mentored junior developers in software best practices, agile methodologies, and version control.
Groovy on Grails Apache Hadoop JavaScript Typeahead.js Jenkins AWS
Jan 2011–Sept 2014

Core Member, Robotics and Embedded Systems

Robotics Club, IOE-WRC, Pokhara
  • Built competition robots from the ground up: designed electronics hardware, wrote control software, and contributed to mechanical fabrication, working day and night in the workshop to meet competition deadlines.
  • Developed image processing systems for intelligent surveillance and criminal detection, integrating computer vision algorithms with real-time embedded hardware.
  • Designed remotely controlled systems using Bluetooth, Wireless 802.11, and RF protocols.
  • Presented and trained at national and college-level workshops on automation, robotics, and emerging technologies.
Embedded Systems Image Processing RF / Bluetooth Circuit Design Arduino
Skills

Technical stack.

Computer Science / ML

Deep learning, CNNs, GANs, Transformers, reconstruction models, privacy-preserving ML, adversarial robustness, and classical ML.

PyTorchTensorFlowCNNs / GANsTransformersLensless ImagingAdversarial Robustness

Software Engineering

Backend development, Java frameworks, cloud computing, experiment automation, and scalable code organization.

JavaPythonScalaSpring BootREST APIsDockerDevOps / CI-CDGitAgile

Data Systems

Large-scale data processing, search, and analytics infrastructure.

HadoopSparkElasticsearchMySQLETLData PipelinesBig Data Analytics

Hardware / Embedded

Electronics background with hands-on experience in microcontrollers, digital logic, and circuit design.

MicrocontrollersArduinoRaspberry PiDigital LogicFPGACircuit Design
Education

Academic background.

PhD in Computer Science

2021–present

University of Kansas, USA. Research focus: privacy-preserving computer vision and lensless imaging.

BE in Electronics & Communication Engineering

2011–2014

Tribhuvan University, Nepal.

Teaching

Courses & workshops.

Courses

Teaching Assistant

Python Programming · Object-Oriented Programming · Data Structures and Algorithms · Digital Logic

Capstone Supervisor

Software Engineering I · Software Engineering II

Workshops & Training

Technical Workshops

Microcontrollers and Computer Organization · Web Application Frameworks (Groovy on Grails, Spring Boot, Java Servlets) · Machine Learning Fundamentals, Version Control, and Reproducibility

Leadership & Outreach

Community, culture, and service.

Leadership & Outreach

Secretary — International Nepali Literature Society (2024–2026)
President — Nepalese Student Association (2021–2022)
President — Engineering Literature Forum (2012–2014)

Academic service

Teaching assistant across multiple CS courses. Capstone project supervision. Workshop organizer for ML fundamentals and reproducibility.

News

Recent updates.

2025
Scheduled talk at AI4EIC.
2025
Paper accepted to ICMLA — Reconstruction-Free Lensless Image Classification.
2024
Paper accepted to IEEE MASS — Robust Privacy-Preserving Classification for Lensless Images.
2024
Submitted steganography manuscript for review.