About Me

I am a Computer Science & Engineering PhD candidate at UC Santa Cruz, advised by Prof. Katia Obraczka in the Inter-Networking Research Group (i-NRG). My work turns commodity wireless telemetry into practical sensing systems, from LLM-based indoor localization to contactless vital-sign monitoring.

I maintain the University of California's Essential Needs web app ecosystem, including the UC Essential Needs Directory and UC Essential Needs Navigator. I built the directory and AI-powered Navigator with campus partners, from backend services to deployment. The platform serves a UC system that enrolled 301,093 students in fall 2025.

Research & Projects

• Master's thesis: Argus: Attention-Guided Transformers for Scalable Wi-Fi CSI Person Identification. Topic: Passive, device-free person identification from commodity Wi-Fi Channel State Information (CSI) using compact statgram representations and a lightweight decoder-only transformer. Status: Manuscript in submission to an AI conference; Argus arXiv. BayLearn 2026 poster (PDF)
• WiFiGPT → Locaris / WiFi telemetry localization (2025–Present). WiFiGPT: GPT-based indoor localization from heterogeneous WiFi telemetry, including CSI, RSSI, and FTM. Locaris: Compact decoder-only LLM that treats each AP measurement as a token and ingests raw WiFi telemetry without preprocessing; published in IEEE PerCom 2026. Result: Sub-meter RSSI/FTM localization, centimeter-level CSI results in WiFiGPT, and few-shot Locaris adaptation to unseen devices and deployments. IP: WiFiGPT technology disclosure listed as Tech ID 34231 / UC Case 2026-577-0; WiFiGPT provisional patent filed. WiFiGPT arXiv · Locaris arXiv · Locaris IEEE Xplore · PerCom 2026 WiP announcement · UC Tech Transfer
• Pulse-Fi (2024–Present). Focus: Low-cost, contactless WiFi CSI sensing for heart-rate, breathing-rate, and apnea monitoring. Publication: Peer-reviewed DCOSS-IoT paper plus extended cardiopulmonary arXiv version; IEEE Xplore listed 3,161 full-text views as of May 20, 2026. IP: Utility provisional patent application filed under 35 USC 111(b). Commercialization: Completed regional I-Corps customer-discovery training through UCSC's Innovation & Business Engagement Hub for Pulse-Fi, including customer interviews, value-proposition testing, and commercialization planning. IEEE DOI · paper PDF · extended arXiv · UCSC News · EurekAlert · Edge AI presentation · press compilation
  Pulse-Fi demo with WiFi hardware and live signal trace. KSBW TV coverage media for the Pulse-Fi paper. Edge AI and Vision Alliance presentation on Pulse-Fi. Pulse-Fi UCSC clinical visit with Jody Rawles, UCI physician. Pulse-Fi site visit with Brian C. Ayers, MD at Harvard/MGH in Boston.
  Pulse-Fi coverage spans 100+ public articles, videos, reposts, and discussions, including: Live Science, CNET, Tom's Hardware, IEEE Spectrum, Medical Xpress, News-Medical, Healthcare in Europe, Interesting Engineering, Hackster.io, ZDNet, XDA Developers, Hackaday, KSBW TV, and Hacker News discussion.
• UC Essential Needs Directory & Navigator (2021–Present). Focus: UC-wide web app for food, housing, health, transportation, legal, and financial-aid support. Build: AI-powered Navigator and complete technical system with UC Essential Needs Consortium, CEJA, and UCSC partners. Admin Console: I maintain the UC Basic Needs admin panel for CRUD updates, public-site feedback review, and Ask AI feedback used to improve answer quality. CEJA Navigator page · admin console · UCEN iOS app · UCSC launch story · CHEBNA 2026 presentation (PDF) · UC enrollment record
• Network Simulation Bridge (NSB). Focus: Open-source co-simulation bridge between front-end apps and OMNeT++ / ns-3 network simulators. Evidence: ACM Q2SWinet 2023 paper, IEEE ICDCS 2025 tutorial, and NSF POSE Award 2449377. ACM DOI · GitHub · ICDCS tutorial · NSF award

Experience

• ML/AI Intern, Zscaler (Summer 2026). Manager: Dat Nguyen. Platform: Deployed LLM services to production for Zscaler's internal LiteLLM gateway using Argo CD, Helm charts, OpenLens, and Prisma Migrate. Managed container images with Amazon Elastic Container Registry (Amazon ECR) and deployments across Amazon Elastic Container Service (Amazon ECS) and Amazon Elastic Kubernetes Service (Amazon EKS), applying Kubernetes sidecar containers and Argo CD's App of Apps pattern. Optimization: Benchmarked LiteLLM Auto Router, Headroom, RTK (Rust Token Killer), and pxpipe optical context compression on Terminal-Bench 2.1 via the Harbor framework. Built a Pydantic-based evaluation pipeline to compare end-to-end cost and task quality across approaches. Impact: Selected and deployed an optimization projected to reduce LLM costs by at least 20% while maintaining task quality.
  With manager Dat Nguyen after presenting the LLM inference optimization project. Zscaler intern cohort during a Ways of Working session. Zscaler intern cohort dinner in San Jose. Rage room outing with fellow interns.
• LLM & Platform Engineer, UC Santa Cruz (Oct 2021–Present). Built and deployed the UC Essential Needs Directory and Navigator with student, staff, and faculty partners. UCSC launch story.
• Graduate Researcher, i-NRG Lab (Sep 2021–Present). WiFi sensing, indoor localization, edge ML, and network co-simulation on commodity hardware.
• Teaching Assistant, UCSC (Jan 2022–Jun 2024). ASTR 119, CSE 130, CSE 13S, CSE 30, CSE 20.

Mentoring

I currently mentor Raksshana Harish Babu (undergraduate, UCSC) on WiFi sensing research. I also mentor Makela Shen (Saratoga High School). Previously mentored Russell Elliott (M.S., UCSC), now a co-author on Locaris.

Since 2024 I have mentored high-school researchers through the UCSC Science Internship Program (SIP).

• SIP 2025 — CSE-07: Enhancing Wi-Fi Sensing Using Machine Learning Models with the 802.11 protocol. Mentor for Ruoyi (Gloria) Tang, William Boyd Nzive, Vir Choudhary, Yuhan (Amy) Li, and Dhriti Agarwal. Yuhan (Amy) Li is continuing on a potential new WiFi sensing project. Slides.
  SIP 2025 CSE-07 WiFi sensing student group.
• SIP 2024 — CSE-15: Enhancing Indoor Localization Using Machine Learning Models with the 802.11 Protocol. Mentor for Pranay Kocheta, Axel Edin, Rafael Castro, and Ella Li. Pranay Kocheta is actively working on Pulse-Fi, Locaris, and future WiFi sensing research. Slides · Project video.
  SIP 2024 CSE-15 indoor localization student group.

Awards, Service & Recognition

• QB3-UCSC Graduate Fellowship for Innovators, 2026 (PulseFi). Selected as one of a few UCSC graduate students receiving GSR support for research poised for commercialization (program page).
• Biodiversity Seed Award ($2,500, 2021–22) and Agroecology Award ($3,000, 2022–23) — UC Basic Needs.
• UC Basic Needs honorarium ($5,000, 2023) and UC Essential Needs Directory & Navigator fellowship ($7,500, 2025).
• CITRIS Smart Health Masterclass & Intensive (2025). Nominated for a selective 5-day program at UC Berkeley on responsible AI in healthcare and healthtech innovation.
• Certificate of Recognition, UC Basic Needs & Economic Justice Center, 2024 and 2025.
• Invited presenter, UC Basic Needs convenings (2022–2026): CHEBNA 2026, CHEBNA 2024, UC GPC Summit 2024 at UCLA, and UC Basic Needs Convening at UC Irvine.
  UC Basic Needs web application workshop at UC Irvine. UC Basic Needs workshop team media at UC Irvine.
• TPC Member, IEEE LCN 2026; Reviewer: IEEE Transactions on Mobile Computing — two manuscript reviews (one in 2025 and one in 2026), ACM TOMPECS (2026).
• Invited talk: Edge AI and Vision Alliance on non-contact vital-sign monitoring with low-cost WiFi devices.
• Young Gladiators / Colosseum Master Class (DARPA/NSF): LinkedIn post · program page.
  Sharing a three-minute wireless research idea with Colosseum researchers at Northeastern.
• Project judge, CruzHacks 2026 (LinkedIn post).

Technical Skills

• Programming: Python, SQL, C/C++, Java, Scala, Solidity
• ML/AI: PyTorch, TensorFlow, Keras, scikit-learn, NumPy/Pandas, Matplotlib, LiteLLM, Pydantic, LLM inference cost optimization, model routing, context compression
• Web: Node.js, React, Express, GraphQL, Flask, Django, HTML/CSS/JS
• Data/Infra: MySQL, MongoDB, Prisma ORM, Prisma Migrate, CUDA, Docker, Kubernetes, Helm, Argo CD, OpenLens, Nginx, AWS (Amazon ECR, Amazon ECS, Amazon EKS), GCP
• MLOps/Automation: Terminal-Bench 2.1, Harbor, GitHub Actions, CI/CD, Jenkins, Ansible, cron, regression testing, Pytest, PyTorch Lightning, Selenium, SLURM
• Cloud Patterns: GitOps, sidecar containers, App of Apps pattern

Earlier — Hackathons, Competitions & Leadership

• NFT-Powered Resume Verification — Winner, Descope Challenge #2 at DeveloperWeek CloudX 2023 Hack. Built a Flask/React/Web3 platform that verifies resume claims using NFT transaction and token IDs, with smart-contract validation, PDF metadata embedding, and ATS-oriented authenticity scoring.
• Winner, Mastek Project Deep Blue Season 5 (2019–2020). Undergraduate team leader; built a visual-acuity testing app that used periscope reflections to virtually increase eye-chart distance, making accurate vision screening possible in small rooms. Winning video · Project video.
• Optimystic MVA. Measure Visual Acuity iOS app. The App Store release did not move forward because publishing a healthcare app required organizational support, while the fees and review process were too burdensome for this project. I decided to keep it open source instead. Feel free to try it out and compile it yourself. Academic project report (PDF) · Original project abstract (PDF)
• Mastek Project Deep Blue Season 6 — Automation Society Security Task. Integrated visitor identity logging with timestamps and temperature checks; demonstrated a gait-recognition model using OpenPose and KNN with MLX90614 temperature sensing. Abstract and slides. This was also where I first came across WiFi CSI and became obsessed enough to make it my PhD research direction.
• Crop Yield Estimation (NDVI + OpenCV). Estimated crop health from drone imagery using YOLOv3 object detection, NDVI vegetation-index calculation, and crop masking toward a Mask R-CNN-style segmentation pipeline. Video · Abstract · Final presentation.
  CDAC AI Hackathon Pune crop-yield estimation project.
• Winner, Smart India Hackathon 2019 (Software Edition). Backend developer on Team Random-6; built "Finding the Nearest Parking Lot in Vicinity."
  Smart India Hackathon 2019 award media. Smart India Hackathon 2019 team media.
• 1st prize (Societal Category), KJSIEIT-IIC IoT Idea Competition 2019 for "Vehicular Ad-Hoc Network and IoT." Project video.
• Chairperson, IoT Cell, KJSIEIT (2020–2021). Organized events and seminars attended by 100+ students.
• Teaching Instructor, Python & ML — KJSIEIT & Connect Club (2021).

Outside Work

Nerd Stuff

• Wi-Fi CSI hardware. Around 2020, very few commodity Wi-Fi chips exposed CSI cleanly, so we tracked down an Intel 5300 NIC and installed it locally to start collecting CSI. More recently, I have also worked with Intel AX210 capture pipelines using FeitCSI.

Contact