Open to full-time roles · Ph.D. expected May 2027

Deep Pujara

Signal Processing & Embedded ML Engineer

Ph.D. Researcher, Electrical Engineering · Arizona State University

I build audio, vision and sensor ML that runs in real time on constrained hardware.

Audio & Speech ML Embedded ML / TinyML Model Compression DSP Computer Vision
Download résumé Contact me
LinkedIn Google Scholar IEEE Xplore Web of Science GitHub
Portrait of Deep Pujara
I go after hard problems, never stop learning, and love building what I haven't built before — then sharing this knowledge to the next generation of engineers.
ResearcherEngineerEducator
12
Publications & manuscripts
59
Citations · h-index 5
27+
Peer reviews · 6 venues
2
U.S. patent applications in progress
3
Internships at Skyworks Solutions
5
Honors & awards
19
Talks, incl. 2 invited lectures
15+
Student researchers mentored
Citation data from Google Scholar, September 2026. Documentation index for reviewers
Experience

Industry & research experience

PythonPyTorchTensorFlow Lite MicroCMATLAB
May – Aug 2025 · May – Aug 2026

AI Speech and Signal Processing Intern

Skyworks Solutions · Hillsboro, OR

  • Built a cycle-accurate Python simulator of a systolic-array ML accelerator (configurable dataflow, clock, memory) for per-layer cycle-count, PE-utilization and bandwidth analysis, reaching a 2–3× inference runtime speedup across candidate designs.
  • Diagnosed a recurring bfloat16 rounding error from reduced mantissa precision vs. FP32 by tracing accumulation errors through the audio inference pipeline.
  • Added auto-generated convolution edge-case tests and structured simulation logging, showing how clock frequency and memory size affect performance.
  • Profiled SRAM/DRAM access patterns and tiling strategies in memory-hierarchy co-design, and deployed audio noise-separation models on Skyworks SoC/ASIC hardware with power measurement.
Audio MLSystolic arraysHW/SW co-designNumerics
Jan 2024 – Present

Graduate Research Associate — Ph.D. Research

SenSIP Lab, Arizona State University · Tempe, AZ

  • Real-time PV array topology classification under partial shading: 93.8% accuracy on a local machine and 90.9% on INT8 hardware, after compressing the network 74% (694 KB → 181 KB) with QAT and PTQ for an Arduino Nano 33 BLE Sense.
  • 90% accuracy from a 47 KB dilated CNN for embedded fault detection (soiling, partial shading) using pruning and INT8 post-training quantization.
  • Designed a custom PCB with XBee wireless mesh that streams 4 real-time sensor channels across a distributed PV monitoring network.
  • Now training a Vision Transformer on solar-array imagery and evaluating on-device inference on an NVIDIA Jetson Nano.
Embedded MLQuantizationComputer visionPCB design
May – Aug 2023

Broadcast Application Engineering Intern

Skyworks Solutions · Austin, TX

  • Delivered a production-ready USB-to-SPI bridge (Rev 2.0) in OrCAD with 4 chip-select/reset lines, improving signal integrity with full drop-in compatibility with Rev 1.0.
  • Wrote low-level C drivers for real-time data transfer across all 4 chip-select channels.
COrCADSPI / USB
Sep 2021 – Dec 2023

Graduate Research Associate — Master's Research

SenSIP Lab, Arizona State University · Tempe, AZ

  • Deployed an embedded ML model on the Arduino Nano 33 BLE Sense (TensorFlow Lite Micro) with 86% test accuracy for real-time solar fault detection, reducing model size by 87%.
  • Designed monitoring hardware integrating voltage, current, temperature and irradiance sensors with 1-second transmission.
TensorFlow LiteArduinoIoT
Earlier research internships — ISRO and Universidad Pública de Navarra (2020–2021)
Jan 2021 – May 2021

Student Research Intern

Indian Space Research Organisation (ISRO) · Ahmedabad, India

  • Developed an algorithm to blindly identify carrier frequency, bandwidth and modulation scheme at 0 dB SNR or higher, with ±1 kHz accuracy in carrier frequency and bandwidth.
  • Surveyed blind-identification and cyclic-spectrum methods, and built two MATLAB applications for modulated-signal generation and feature estimation.
Signal processingWireless communicationsMATLAB
Jun 2020 – Aug 2020

Student Research Intern

Universidad Pública de Navarra · Pamplona, Spain (virtual)

  • Developed a MATLAB algorithm to transform near-field antenna radiation patterns to far-field using interpolation and probe correction.
  • Designed a radio link-budget calculator for terrestrial and satellite links, including atmospheric absorption by oxygen and water vapor.
AntennasRF propagationMATLAB
Research

Reliable ML on constrained hardware

Four connected research threads, with measured results.

Embedded ML for solar PV

Fault detection and array reconfiguration running on microcontrollers.

TinyMLQAT / PTQ

Audio & speech

On-device keyword spotting and neural speech enhancement.

MFCC / STFTGRU

ML hardware & numerics

Accelerator simulation, bfloat16 error analysis, GPU vs. Gaudi benchmarks.

Systolic arraysbfloat16

Education & quantum ML

Quantum AI tools in signal-processing curricula and NSF programs.

IEEE FIEICASSP

Why it matters

The practical value of running ML at the edge.

More reliable solar energy

Detecting faults and shading on-site, in real time, lets PV arrays be monitored and reconfigured without cloud connectivity or costly equipment.

Energy-efficient AI

Compressing models to kilobytes lets audio and sensing AI run on low-power chips instead of servers — the same problem behind hearing devices, wearables and IoT.

Engineering workforce

Teaching deep learning and signal processing at ASU and mentoring students through NSF-funded research programs.

Key results

Solar PV
93.8% / 90.9%
PV topology classification
Local machine / INT8 on Arduino Nano 33 BLE Sense
Solar PV
74%
Model compression
694 KB → 181 KB with QAT and PTQ
Solar PV
90% @ 47 KB
Embedded fault detection
Dilated CNN, pruning + INT8
Audio
91.5%
Keyword spotting
9K-parameter BC-ResNet, 10 keywords, 4× smaller with INT8
Audio
2.56 PESQ
Speech enhancement
0.937 STOI on 824 VoiceBank+DEMAND test utterances
Hardware
2–3×
Accelerator runtime speedup
Cycle-accurate systolic-array simulator (Skyworks)

Selected projects

Aug 2025 – Oct 2025

Neural Speech Enhancement Using Recurrent Masking

  • GRU predicts per-frequency-bin gains from 40-dim MFCC features: 2.56 PESQ, 0.937 STOI on 824 held-out VoiceBank+DEMAND utterances.
  • 4-stage pipeline: STFT ideal ratio masks, phase-preserving reconstruction, variable-length masked training, automated PESQ/STOI evaluation.
PyTorchGRUSTFT
Jul 2024 – Sep 2024

Efficient Keyword Spotting on Microcontrollers (BC-ResNet)

  • 91.5% accuracy across 10 keywords with a 9K-parameter BC-ResNet (2D frequency, dilated 1D temporal and pointwise convolutions).
  • 4× smaller with INT8 TensorFlow Lite; runs fully on-device on an Arduino.
TensorFlow Lite MicroINT8Arduino
Publications

Publications & patents

Google Scholar
12
Publications & manuscripts
59
Citations
6
First-author papers
5
h-index · i10-index 3
Solar & embedded ML Education & quantum ML IoT & sensing
2026

Real-Time Photovoltaic Array Topology Optimization Using Embedded Machine Learning

D. Pujara, D. Srinivasan, C. Tepedelenlioglu, A. Spanias

IEEE Access (journal) — submitted

Under reviewIEEE Access
2026

Embedded Hardware and Edge ML Algorithms for Real-time PV Fault Detection

D. Pujara, M. Hansen, N. A. Babar, D. Srinivasan, C. Tepedelenlioglu, A. Spanias

International Conference on Information, Intelligence, Systems & Applications (IISA), Greece, Jul. 2026

AcceptedIISA
2024

Real-time PV Fault Detection using Embedded Machine Learning

D. Pujara, D. Ramirez, C. Tepedelenlioglu, D. Srinivasan, A. Spanias

2024 IEEE 7th Int. Conf. on Industrial Cyber-Physical Systems (ICPS), St. Louis, MO, pp. 1–5

Bagging ensemble on an Arduino Nano: 85.97% test accuracy across four PV fault conditions, on-device.

First authorIEEE ICPS
Cited by 10
2023

Design of a New Photovoltaic Intelligent Monitoring and Control Device

D. Pujara, D. Ramirez, C. Tepedelenlioglu, D. Srinivasan, A. Spanias

2023 14th Int. Conf. on Information, Intelligence, Systems & Applications (IISA), Volos, Greece, pp. 1–4

Sensor hardware for voltage, current, temperature and irradiance with 1-second transmission.

First authorIEEE IISA
Cited by 3

Patent applications 2

Real-Time Photovoltaic Array Topology Reconfiguration Using Embedded Machine Learning

D. Pujara, et al. · U.S. patent application — filing process initiated, 2026

Embedded Hardware and Edge ML Algorithms for Real-time PV Fault Detection

D. Pujara, et al. · U.S. patent application — filing process initiated, 2026

Teaching & education

Teaching, mentoring & degrees

4
ASU courses as Teaching Associate
200+
Students supported
15+
Undergraduate researchers mentored
3
Lecture modules authored
Aug – Dec 2025 · Aug 2026 – Present

Graduate Teaching Associate

EEE 598 Deep Learning & EEE 515 Computer Vision · ASU

  • Mentor graduate students through 4 architecture families (CNN, RNN, Transformer/ViT, GAN) in full PyTorch pipelines; led 14 class sessions on DL architectures.
  • Benchmarked training on Intel Gaudi 2 vs. NVIDIA A100 on ASU's Sol supercomputer for course experiments.
Jan 2023 – May 2025

Graduate Teaching Associate

EEE 407 Digital Signal Processing & EEE 606 Adaptive Signal Processing · ASU

  • Supported 200+ students in FFT, filter design and sampling theory; led 2–3 problem-solving sessions per semester.
  • Authored and recorded 3 lecture modules (LLMs, Transformers, Attention) for the graduate signal-processing curriculum.
2022 – 2025

NSF REU / RET Research Mentor

SenSIP Lab, Arizona State University

  • Mentored 15+ undergraduate researchers and K-12 teachers in MATLAB, Python and ML through NSF-funded programs, from literature review to publication-quality results.
2021 – 2022

Graduate Grader

EEE 407 Digital Signal Processing · ASU

  • Graded assignments and homework for the combined undergraduate–graduate DSP course.

Education

Arizona State University
Jan 2024 – May 2027 (expected)

Ph.D., Electrical Engineering

Arizona State University

GPA 3.83 / 4.00

Signal processing · embedded ML · solar energy

Arizona State University
Aug 2021 – Dec 2023

M.S., Electrical Engineering

Arizona State University

GPA 3.82 / 4.00

Signal processing · ML · solar energy

Nirma University
Aug 2017 – May 2021

B.Tech., Electronics & Communication

Nirma University, Ahmedabad, India

GPA 8.11 / 10.00

Graduate coursework

Digital Signal ProcessingAdaptive Speech ProcessingDeep LearningComputer VisionArtificial Neural ComputationRandom Signal Theory
Professional service

Peer review & memberships

27+ manuscript reviews for 6 journals and conferences, with certificates or verified records linked.

Verified reviews · Web of Science
IEEE Access

IEEE Access

IEEE · Journal · ML & signal processing

10 reviews
Web of Science
IEEE FIE

IEEE Frontiers in Education (FIE)

IEEE · Conference · 2025 & 2026

6 reviews
2025 certificate2026 certificate pending
IEEE

IEEE AIMV 2025

IEEE · Conference on AI & Machine Vision

5 reviews
Certificate
Elsevier

Solar Energy

Elsevier · Journal · Solar energy & ML

4 reviews
Certificate
IEEE

IEEE Transactions on Technology and Society

IEEE · Journal

2 reviews
Web of Science
Grace Hopper Celebration

Grace Hopper Conference

AnitaB.org · Conference · Women in computing

Reviewer

Honor societies

IEEE-HKN Epsilon Beta chapter induction ceremony at Arizona State University
IEEE-HKN

IEEE Eta Kappa Nu (IEEE-HKN)

Inducted into the Epsilon Beta chapter at Arizona State University. IEEE-HKN is the honor society of IEEE, recognizing excellence in electrical and computer engineering.

Inducted Apr 2026Induction photo
Sigma Xi Full Member certificate
Sigma Xi

Sigma Xi, The Scientific Research Honor Society

Elected a Full Member by the Committee on Qualifications and Membership. Sigma Xi honors scientific research achievement across disciplines.

Full Member · 2026Certificate
Honors & awards

Recognition

Renew Challenge 2023 certificate
Winner · Team Funded2023

Renew Challenge 2023

12th Arizona Student Energy Conference

Six-member team proposal selected for funding (“RENEW Challenge Team Funded”) for renewable-energy and sustainable technology.

Renew Challenge 2024 certificate
Winner · Team Funded2024

Renew Challenge 2024

13th Arizona Student Energy Conference

Selected for funding for the second consecutive year with an advanced renewable-energy project proposal.

Smart India Hackathon 2020 winner certificate
Winner2020

Smart India Hackathon 2020

ISRO problem statement

Led the team to a top-6 placement among 50+ teams on ISRO space-communication problem statements.

News article
Best Paper certificate, ACECAT
2nd Best Paper2021

Best Paper Award — ACECAT

Conference on Advancement in Communication, Electronics, Computer & Automation Technology

“Blind Scrambling Code Identification for Synchronous and Self-synchronous Scrambler.”

Proceedings (p. 75)
SenSIP certificate
Certificate

SenSIP Certificate of Excellence

Arizona State University

Completed background and application-oriented training in Sensors, Signal and Information Processing.

Certificate program
Talks & presentations

Speaking

2 invited lectures and 17 presentations. Each card links to the original post.

Invited talk at Adani University
Invited lecture · Jan 2026

How Humans Taught Machines to Think: A Journey of AI

Adani University, Ahmedabad, India

 Invited talk at Nirma University
Invited lecture · Dec 2025

Generative Pre-trained Transformers — A Combination of a Few Individual Innovations

Nirma University, Ahmedabad, India
Skills

Technical toolkit

Programming

Python MATLAB C / C++

Audio & signal processing

Digital filteringFFT / STFTSampling theoryMFCCAdaptive filteringFixed-point inference

Model compression

PruningQuantization-aware trainingINT8 post-training quantization

ML frameworks & data

PyTorch TensorFlow TF Lite Micro Scikit-learn NumPy Pandas Matplotlib

Embedded hardware

Arduino Nano 33 BLE Sense NVIDIA Jetson Orin NanoRaspberry Pi HAT+ 2ESP32XBee S2COrCAD PCB

Tools

Git VS Code Netron JIRA Confluence Simulink LaTeX
Recommendations

What colleagues and mentors say

8 recommendations from industry engineers, a professor and students, published on LinkedIn.

View on LinkedIn
Deep was an excellent addition to our team over the Summer of 2025. We spent countless hours whiteboarding and then prototyping deep learning inference optimizations. He made substantive contributions to our platform and to the ideas we are exploring. His ability to quickly understand complex ideas and turn out prototypes was impressive.
Skyworks

Spencer Hann

Senior Deep Learning Engineer, Skyworks Solutions

Industry
I found Deep to be an excellent team player, self-starter, great communicator of ideas, and precise scientific thinker about and quick implementer of highly complex ideas. He has made extremely valuable contributions to our team. I highly recommend him!
Skyworks

Alexander Kain

Sr Principal AI Systems Engineer, Skyworks Solutions

Industry
My interaction with Deep has been during his BTech study at Nirma University. He did a number of academic projects with me. During the project work, Deep demonstrated the ability to work independently with great creativity and enthusiasm. Deep is well organized, diligent, and a fast learner. He is one of the most dedicated student that I've worked with and is always willing to put extra efforts in all of the work assigned to him. I wish him success for all his endeavors.
Nirma University

Dr. Sachin Gajjar

Associate Professor, Nirma University

Professor
During bachelor's program, Deep worked under my guidance for several projects, specifically in the domain of EM & antennas. I found him dedicated and sincere towards the work. I have noticed a good quality of proactiveness while working with him in collaboration. He always try to see the problem as a big picture and apply planned approach for the successful closure. I wish him a great success and happy to recommend his name for any future references.
Intel

Jay Gupta

Sr Antenna Engineer, Intel Corporation

Project guide
I had the pleasure of working with Deep in the same research lab during my master's thesis. While our projects were different, we often brainstormed together, and I was always impressed by his ability to tackle complex problems with creative solutions. He is hardworking, passionate, and an excellent researcher. I strongly recommend him for any future opportunities.
onsemi

Bhavikkumar Patel

Product Engineer II, onsemi

Lab colleague
During my time as an undergraduate research intern at ASU-SenSIP, Deep Pujara was my mentor for a project in machine learning. He provided me with a solid foundation in classical machine learning, sharing his knowledge and expertise to help me quickly get up to speed on the background research. His leadership and technical guidance created a highly productive and collaborative environment. I am pleased to recommend Deep for his professionalism and dedication to fostering the growth of his mentees.
ASU

Diego Reyes

Sr Computer Science Student, Arizona State University

Mentee
Deep is one of the strongest technologist I have worked with. He has a quick grasping power, excellent deep dive and troubleshooting ability. He is a fast learner and even as an undergraduate student he was involved in some really challenging and innovative projects related to communications and programming. His enthusiasm is contagious. It was a pleasure to work with him and brainstorm any new idea.
Crest Data

Praveen Kukreja

Sr Software Engineer, Crest Data

Colleague
Deep is knowledgeable, articulate, and innovative person to work with. I have known him for quite a time during our educational journey and being a colleague and friend of him, he is someone I trusted and always looked forward to work with him. I have seen him closely working on different projects and have always analyzed his ability to bring integrity, innovation and intelligence to his work and also his managing skills. I believe his overall presence positively impacted the work and the people around him.
IBM

Dhruv Shah

Backend Developer, IBM

Colleague
For reviewers

Documentation index

Public sources for the achievements on this page. More documentation is available on request.

CategoryItemDateSource
RésuméCurrent résumé (PDF)Sep 2026Deep_Pujara_Resume.pdf
PublicationsCitation record: 59 citations, h-index 5Sep 2026Google Scholar
PublicationsIEEE author profile (published papers)2020 – 2026IEEE Xplore
Peer reviewVerified review record — IEEE Access (10), IEEE Transactions on Technology and Society (2)2025 – 2026Web of Science
Peer reviewReviewer certificate — IEEE FIE 2025Nov 2025Certificate (PDF)
Peer reviewReviewer certificate — IEEE AIMV 2025Aug 2025Certificate (PDF)
Peer reviewCertificate of Reviewing — Elsevier Solar Energy (4 reviews, May–Sep 2025)Sep 2025Certificate (PDF) · Post
Peer reviewReviewer — IEEE FIE 2026 (3 reviews)2026Certificate pending
MembershipSigma Xi — elected Full Member2026Certificate (PDF)
MembershipIEEE-HKN — Epsilon Beta chapter induction, ASUApr 2026Induction photo
AwardSmart India Hackathon 2020 — winner certificateAug 2020Certificate · News
Award2nd Best Paper — ACECATFeb 2021Certificate · Proceedings p. 75
AwardRenew Challenge 2023 and 2024 — “Team Funded” certificatesApr 2023 · Apr 20242023 · 2024
TalksInvited lectures, Adani University and Nirma UniversityJan 2026 · Dec 2025Adani · Nirma
Letters8 recommendations (industry, academic, peers)—LinkedIn
Contact

Let's talk

Open to full-time roles in audio ML, signal processing and embedded ML, and to research collaborations.

University email
dpujara1@asu.edu
Location
Tempe, Arizona, USA
LinkedIn Scholar GitHub