Duong Anh (Tom) Nguyen
CS student · robotics · ML · building things from first principles
About
I'm Duong Anh (Tom) Nguyen, a computer science student at San Francisco Bay University. I care about robotics, machine learning, and how AI is built and deployed in the real world.
Right now I'm an AI engineer and founding member at Quiche (ADU Marketplace), a student assistant in the university IT department, and president of the Robotics Club (2026).
Previously I interned on RAG and LLM projects—shipping internal AI systems end to end for a business setting.
I'm writing a paper on the ethics of using AI to restore or alter historical war imagery: what we gain, what we risk, and who decides.
On the side I'm building TTorch, a small PyTorch-style deep learning framework in C++ from scratch, to understand training and autograd at the metal level—not just the API.
Long term I want to work where robotics, ML, and responsible AI overlap. I'm especially interested in AI ethics from an engineer's perspective: clear tradeoffs, not slogans.
I attended ICPC Pacific Northwest Regional Contest 2025 Division 2
I learn by building, think by writing, and try to share what I figure out along the way.
Stay foolish, stay hungry - Steve Jobs
The difficulty exists because you have the capability to solve it
Experience
Nov 2025 — Present
Founding Member & AI Engineer - AI DevOps & ML
Quiche, Inc. (ADU Marketplace)
Built an MCP (Model Context Protocol) server so product features can pull the right context on demand. Designed a multi-agent stack for a smart assistant—roles, orchestration, and retrieval—and stood up Google Cloud (Compute Engine, Cloud Run, Cloud SQL) for deployments that can scale in production.
MCP · Python · Multi-agent AI · Google Cloud · Cloud Run · Cloud SQL
Sep 2025 — Present
Student Assistant, IT Department
San Francisco Bay University
First-line support for hardware, software, and network issues—triage tickets and fix what we can on the spot so classes and staff stay online. Helped track IT assets across campus so deployments, swaps, and maintenance stay visible.
IT support · Networking · Hardware · Asset management · Help desk
May 2022 — Aug 2025
LLM Intern — Implementation
Sofeast Ltd.
Interned on internal RAG and LLM systems: retrieval, prompting, and wiring models into real workflows. Shipped tooling around evaluation and deployment so teams could iterate without losing track of quality.
RAG · LLM · Python · FAISS · OpenAI API · LLaMA · Qwen · Gemma · DeepSeek · GPT-4
Education
Sep 2024 — Dec 2027
B.S. Computer Science
San Francisco Bay University
Foundations in algorithms, systems, and mathematics. Dean's list
Algorithms · Systems · Mathematics · ICPC · Robotics · Machine Learning · Deep Learning · AI · Computer Vision · Natural Language Processing · Robotics · Computer Science
Sep 2021 — Jun 2024
High School
Le Hong Phong High School for the Gifted - Nam Dinh Province
Class of 2024 - Major in Chemistry
Chemistry · Chemistry Competition · Chemistry Education · Chemistry Experiment · Chemistry Presentation · Chemistry Report
Projects
A lightweight tensor and automatic differentiation library written in C++ with Python bindings. It explores the internal architecture of modern deep learning frameworks by implementing tensor operations, an autograd engine, and modular compute kernels, with the goal of building a minimal educational backend similar to PyTorch.
C++ · Python · Autograd · Deep Learning · Systems · PyTorch · Vectorized Autograd
An agentic virtual assistant that uses LLMs as a core reasoning engine combined with Siri-like permission-based action execution. Built on a delegation architecture with an orchestrator agent routing tasks to specialized sub-agents, the system emphasizes security, ethical execution, and real-world action capabilities beyond traditional chatbots.
LLM · Agents · AI Assistant · Pydantic · Voice Interface · Python · FastAPI · Docker · PostgreSQL · GitHub · Git · GitHub CI/CD · Delegation Architecture · Orchestrator Agent · Sub-agents · Ethical Execution · Real-world Action Capabilities · OpenAI API · LLM · Agents · AI Assistant · Pydantic · Voice Interface · Python · FastAPI · Docker · PostgreSQL · GitHub · Git · GitHub CI/CD · Delegation Architecture · Orchestrator Agent · Sub-agents · Ethical Execution · Real-world Action Capabilities
A two-stage real-time emotion recognition system that detects faces from a live webcam feed using YOLOv11n and classifies one of seven emotions with a custom-trained logistic regression model. The pipeline uses ViT-B/16 for 768-dimensional feature extraction and a from-scratch PyTorch classifier trained with L-BFGS, with manual evaluation metrics and a fully modular training and deployment workflow.
Computer Vision · PyTorch · YOLO · ViT · Logistic Regression · Real-time Emotion Detection · Webcam Feed · Face Detection · Emotion Classification · Custom-trained Model · L-BFGS · Manual Evaluation Metrics · Modular Training and Deployment Workflow · Transfer Learning
An interactive maze-solving visualization tool that demonstrates and compares pathfinding algorithms in real time. Users can choose from multiple maze layouts and watch BFS, Flood Fill, and a randomized search algorithm explore and solve them step by step, making it easy to compare performance, efficiency, and solution paths.
Algorithms · Python · React · JavaScript · Visualization
Writing
Draft / ongoing. Compares fully automated, hybrid (AI plus human review), and non-AI restoration for damaged or souvenir war images. We connect capability to authenticity, trust, and collective memory; surface risks such as biased training data and invented detail; and argue—treating photos as historical evidence—that ethical practice needs transparency, oversight, and institutional guardrails. Includes a checklist for spotting historical hallucination in outputs and early results from a 50-person vignette study on trust and acceptability among archivists, families, and online audiences.
Draft / ongoing. Explores Rapidly-Exploring Random Trees (RRT) for maze navigation when state and controls are represented in a vector form suited to continuous spaces and simple collision checks. Motivation, algorithm sketch, and how the vectorized setup affects sampling, expansion, and path quality compared to a grid-first mindset.
Reading
- BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding— Devlin et al.[Reading]
The BERT paper. Key insight: pre-training a model on a large corpus of text and then fine-tuning it on a specific task.
- GRE Pre Plus 2024— ETS[Reading]
Currently on GRE Pre Plus 2024. Excellent reference for GRE preparation.
The foundational transformer paper. Key insight: self-attention can replace recurrence entirely.
Other
Apr 2026
3rd Place — Wildfire detection & response
Innovate Bay (BayHawk 2.0)
Weekend build: combine live camera fire detection (YOLOv8), NASA FIRMS thermal hotspots, and weather context so the system can corroborate or dismiss alerts, reason with an LLM, label severity from low to critical, and suggest response steps. Implemented with PydanticAI + GPT-4o, async orchestration, and a FastAPI service tying sensing → analysis → action.
Hackathon · Multi-agent AI · YOLOv8 · NASA FIRMS · FastAPI · PydanticAI · GPT-4o · Innovate Bay · BayHawk 2.0 · Verizon Innovation Lab · FastAPI · Python · Vite · Tailwind CSS · PostgreSQL · Docker · GitHub · Git · GitHub CI/CD · Vercel
Mar 2026
Hackathon — KryptosProof
HackHayward (Cal State East Bay)
24-hour build for the “vibe coding” era: an AI red-team that hits your app with real tools (e.g. sqlmap, Nuclei, FFUF), keeps only reproducible findings, drafts and checks fixes, and ships a SHA-256–signed report. Multi-agent design, Docker sandbox, live dashboard, and attestation story—security without hand-wavy noise.
Hackathon · Security · Docker · Nuclei · Multi-agent · HackHayward · KryptosProof
Mar 2026
Winner — Uniflow (AI career path for SFBU)
ThinkNext 2026
Multi-agent career copilot for San Francisco Bay University students: orchestration in PydanticAI with specialized agents for goal intake and resume critique, RAG over courses and events with embeddings + vector search, and a mock interview agent with voice/text (TTS/STT) and STAR-style feedback.
Hackathon · PydanticAI · RAG · OpenAI · TTS · STT · Streamlit · SFBU · Career
Oct 2025
2nd Place — Bit for Bit (receipts → inventory)
Innovate Bay (BayHawk 1.0)
BayHack 2025 (business-action track): capture receipts from a phone camera, extract structured fields into MongoDB, and layer an assistant so small teams can reconcile stock without spreadsheet chaos. Pitched the solution live with teammates. Stack: OpenAI API, MongoDB, Python, Node.js, FastAPI.
Hackathon · BayHack · OpenAI API · MongoDB · FastAPI · Node.js · Python · OCR · Inventory
Feb 2025
Top 3 — BleuLaMuse (career chatbot)
SFBU IEEE — Mind Sparks
IEEE campus competition: conversational career guidance on top of structured student data, with FAISS-backed retrieval for course and pathway suggestions, Streamlit for the interface, and prompt tuning on domain-specific material. Designed to stay modular before a wider cloud rollout.
IEEE · GPT · FAISS · Streamlit · Python · SFBU · Career · Chatbot
Now
Drafting two papers: AI restoration of war photographs (ethics + vignette study) and RRT with a vector maze formulation. Public repos in motion: TTorch (C++/Python autograd), Desir (agentic assistant), a real-time emotion CV stack, and a maze-solving visualizer. Tightening this minimal site to match reality.
Third place at Innovate Bay (BayHawk 2.0) for a wildfire stack: YOLOv8 + NASA FIRMS + weather context, fused with a multi-agent LLM pipeline (PydanticAI, GPT-4o, FastAPI) for severity and response planning.
HackHayward weekend: shipped KryptosProof — multi-agent, tool-using red team (sqlmap / Nuclei / FFUF), Docker sandbox, live dashboard, and signed reports for the vibe-coding security angle.
Won ThinkNext 2026 with Uniflow — a PydanticAI-orchestrated career copilot for SFBU: goal + résumé agents, RAG over courses/events, and a voice/text mock interview path with STAR-style feedback (Streamlit demo).
Joined Quiche (ADU marketplace) as a founding member — AI / DevOps focus: MCP server for contextual retrieval, multi-agent assistant design, and GCP (Compute Engine, Cloud Run, Cloud SQL) for production-shaped deploys.
Contact
Feel free to reach out. I'm always happy to chat about deep learning, open source, or interesting problems.