-6D28D9?style=flat-square&logo=codeforces&logoColor=white){kind=icon}
-4C1D95?style=flat-square&logo=nasa&logoColor=white){kind=icon}
-5B21B6?style=flat-square&logo=starship&logoColor=white){kind=icon}
name: "Ethan Poon"
role: "Software Engineer & AI/ML Engineer"
education: "B.S. Computer Engineering @ Rutgers University Honors College"
location: "Edison, NJ"
focus: ["Distributed Systems", "AI/ML", "Full-Stack Engineering", "IoT Platforms"]I'm a Computer Engineering student at Rutgers University Honors College, building production-grade systems that sit at the intersection of software engineering, distributed infrastructure, and applied AI/ML. My work spans fault-tolerant IoT platforms, real-time stream processing, and multimodal machine learning pipelines — with an emphasis on shipping systems that hold up under real operating conditions, not just in notebooks.
I approach engineering with a product mindset: every pipeline, API, or model I build is evaluated not just on correctness, but on reliability, latency, and the actual workflow it replaces. I care about clean architecture, measurable impact, and systems that fail gracefully.
Currently:
- 🏗️ Architecting a 4-layer fault-tolerant IoT platform for industrial hardware control
- 🧠 Applying computer vision and NLP to real-world diagnostic and geospatial problems
- ⚙️ Designing distributed backend pipelines for large-scale data processing
Languages
Frontend
Backend & Databases
AI / ML
Cloud, DevOps & Tooling
| Domain | Proficiency | Details |
|---|---|---|
| Computer Vision | ⭐⭐⭐⭐⭐ | ConvNeXt CNNs, Grad-CAM, XAI tooling, multimodal medical imaging inference |
| NLP / Speech | ⭐⭐⭐⭐ | Accent & sentiment labeling pipelines, speech-to-text accuracy tuning at scale |
| Geospatial ML | ⭐⭐⭐⭐ | Random Forest land-cover classification, LANDSAT vegetation index integration |
| MLOps | ⭐⭐⭐⭐ | Model validation pipelines, training-run orchestration, dataset preprocessing at scale |
| Self-Supervised Learning | ⭐⭐⭐ | Applied research on representation learning without labeled data |
| Frameworks | ⭐⭐⭐⭐⭐ | PyTorch, TensorFlow Extended, scikit-learn |
🌊 AWMS (Autonomous Water Management System) — Fault-Tolerant IoT Sensor Pipeline
Real-time, end-to-end IoT data pipeline for monitoring industrial water-quality sensors (turbidity, temperature, UV, ATP), designed with store-and-forward buffering to guarantee zero data loss across multi-hour network partitions.
| Stack | Python, Go, MQTT (Mosquitto), Apache Kafka, TimescaleDB, React, Recharts |
| Scale | Multi-trailer sensor network with continuous edge-to-cloud telemetry |
| Performance | Sub-2-second control latency on industrial hardware actuation |
| Security | Isolated edge nodes with buffered local persistence (SQLite) before transmission |
| Impact | Reduced audit prep time by 95%; validated edge hardware in production |
| Repository | PSEGdatapipeline |
Architected as a 4-layer system: a sensor simulator publishes telemetry over MQTT to a Python edge node, which persists locally and forwards to a Go ingestion server. The Go server produces to Kafka, a consumer writes to TimescaleDB, and a React dashboard polls live data with spike-alert detection — built for resilience against real-world network instability on distributed hardware.
🏫 CPS WebApp — Distributed Systems Benchmark Automation
Full-stack application built with the University of Pennsylvania's Distributed Systems Lab to automate the preprocessing of benchmark execution profiles for systems research.
| Stack | React, Node.js, Express, MongoDB, REST APIs |
| Scale | Processes real-time systems execution trace data across benchmark runs |
| Performance | Eliminated 15 hrs/week of manual preprocessing workflows |
| Security | REST API validation layer for incoming trace data |
| Impact | Accelerated system profiling by 40% via DNA/RASCO trace clustering |
| Repository | cps-webapp |
Designed and shipped end-to-end, integrating clustering algorithms to compute resource constraints from execution traces, turning a manual research workflow into an automated pipeline used directly by the lab.
🚢 Mayflower — Congressional App Challenge Winner
Award-winning application built for the NASA App Challenge, recognized as the 1st place submission out of 1,500 entries.
| Stack | JavaScript, HTML/CSS, REST APIs |
| Scale | National-level competition submission |
| Performance | N/A (application-focused, not infra-benchmarked) |
| Security | Standard client-side data handling practices |
| Impact | 1st place out of 1,500 submissions nationally |
| Repository | Mayflower-Congressional-App-Challenge-Winner |
Built end-to-end as a competition submission, combining a clean front-end experience with functional problem-solving logic that stood out among 1,500 nationwide entries.
📈 Stock Prediction — Applied ML Forecasting
Machine learning notebook project applying predictive modeling techniques to historical stock market data.
| Stack | Python, Jupyter, scikit-learn, PyTorch |
| Scale | Historical time-series financial data |
| Performance | Model iteration and evaluation via notebook-based experimentation |
| Security | N/A (research/educational context) |
| Impact | Applied research into time-series forecasting techniques |
| Repository | Stock-Prediction |
Exploratory project applying supervised learning models to financial time-series data, focused on feature engineering and evaluating forecasting accuracy across different modeling approaches.
🧠 QBrainX — Applied AI/ML Systems
Python-based AI/ML systems project exploring applied machine learning techniques and model architecture design.
| Stack | Python, PyTorch, scikit-learn |
| Scale | Research/prototype scale |
| Performance | Iterative model benchmarking |
| Security | N/A (research/educational context) |
| Impact | Hands-on exploration of applied ML system design |
| Repository | QBrainX |
A applied machine learning systems project focused on building and iterating on model architectures in Python, used as a hands-on extension of coursework and independent research.
May 2026 – Present | Newark, NJ
Architecting a fault-tolerant, 4-layer IoT platform for industrial hardware monitoring and control, built for zero data loss across multi-hour network partitions.
- Designed store-and-forward buffering across Raspberry Pi edge nodes, Go services, Kafka, and TimescaleDB
- Built core real-time stream processing features for live field sensors (turbidity, UV, ATP) automating chemical equipment responses
- Achieved sub-2-second control latency on industrial hardware while validating edge devices in production
- Reduced audit prep time by 95% through automated data pipelines
Go Kafka TimescaleDB Raspberry Pi MQTT Distributed Systems
Oct 2025 – Apr 2026 | Philadelphia, PA
Designed and shipped a full-stack application automating benchmark execution profile preprocessing for systems research.
- Built REST APIs to process real-time systems data, eliminating 15 hrs/week of manual workflows
- Integrated DNA and RASCO algorithms to cluster execution traces and compute resource constraints
- Accelerated system profiling throughput by 40%
React Node.js Express MongoDB REST APIs
Sep 2024 – Sep 2025 | Remote
Designed end-to-end backend pipelines for a large-scale multilingual speech corpus (CASPER) used in model training.
- Built REST ingestion, validation, and normalization pipelines for a 1,000-speaker multilingual dataset
- Cut data-prep overhead by 60%, unblocking parallel model training workflows
- Improved speech-to-text accuracy by 15% across 450 training runs
Backend Systems REST APIs Audio Processing Data Pipelines
May 2024 – Sep 2024 | Austin, TX
Engineered scalable geospatial batch pipelines for automated land-cover classification.
- Built Python pipelines (scikit-learn, GDAL) for Random Forest classification across 450+ sq km and 50+ U.S. locations
- Integrated 10+ years of LANDSAT vegetation indices (NDVI, TCG, spectral wavelengths) into the classification model
- Reduced per-site analysis time by 90%; informed agrivoltaic layout decisions projected to increase crop yield by 12%
Python scikit-learn GDAL Geospatial ML Random Forest
Jul 2024 – Oct 2024 | Palo Alto, CA
Engineered a multimodal inference pipeline for automated respiratory condition detection from medical imaging.
- Built a pipeline combining RadGraph and Grad-CAM to process 10,000+ scans from Stanford's CheXpert dataset
- Designed and trained a ConvNeXt CNN with explainable AI (XAI) tooling, achieving 94% diagnostic accuracy
- Model validated and adopted by clinical professionals for diagnostic support
PyTorch Computer Vision Grad-CAM XAI Medical Imaging
Feb 2021 – Dec 2025 | Remote
Founded and led a non-profit delivering free production websites to small businesses and nonprofits through a global volunteer network.
- Architected a standardized HTML/CSS site-delivery system adopted across 132 global chapters
- Enabled 740+ volunteers to ship 650+ production websites
- Secured a $5,000 U.S. Department of Education grant
HTML/CSS Non-Profit Leadership Volunteer Systems Design
| Recognition | Details |
|---|---|
| 🥇 USA Computing Olympiad — Gold | Top 5% internationally |
| 🚀 NASA App Challenge — Winner | 1st place nationally out of 1,500 submissions |
| 🌕 NASA App Challenge — Innovator Award (Artemis) | Best app visualizing Artemis III South Pole landing regions with rover integration |
| 🏛️ Congressional App Challenge — 1st Place | 1st nationally among 2,000+ competitors; presented in Washington, D.C. |
| 💎 Diamond Innovation Challenge — Finalist | 10th place out of 900 submissions |
| 📱 FBLA National Mobile App Development — 3rd | International; competed against 700 teams building a React Native app with AI algorithms |
| 🧪 "Skew the Script" ML Contest — National Winner | 1 of 7 winners out of 9,500 entries; only winning team from NJ |
| 📡 IEEE Grant Recipient | International grant awarded for an Arduino-based engineering project |
| 🎖️ Congressional Silver Medal | 200+ verified volunteer hours across service and development initiatives |
| 🇺🇸 President's Volunteer Service Award — Gold (2x) | Awarded in consecutive years for 100+ and 250+ service hours |
| 🥤 Coca-Cola Scholarship — Semifinalist | Advanced in a highly selective national scholarship program |
| ✍️ National Poetry Publications | Multiple national placements, including Congressional and NomadArtX recognitions |
AWS
Oracle
NPTEL
Cisco
Add/update certification badges above with your verified credential links.
Learning:
- Distributed systems design at scale
- Advanced MLOps and model deployment pipelines
- Systems programming in Go
Building:
- AWMS (Autonomous Water Management System) — fault-tolerant industrial IoT pipeline
- AI-driven diagnostic and geospatial ML tooling
Exploring:
- Self-supervised learning techniques
- Edge computing and store-and-forward architectures
Open To:
- Software Engineering internships
- AI / ML engineering roles
- Research collaborations in distributed systems
"Build systems that survive contact with reality."
