I work directly with non-technical stakeholders—product owners, business analysts, biology researchers—and turn their problems into shipped production systems. Over 2+ years I've built features for an enterprise Vue / C# .NET / SQL Server platform deployed in 100+ countries, delivered custom integrations and data migrations, and deployed AI features including a RAG-based Voice AI answering Vietnamese questions from English PDFs with citations. I graduate in 2026 with a B.S. in Computer Science: Software Engineering from Liberty University.
const loc: Engineer = {
name: "Loc Nguyen",
role: "Software Engineer",
company: "Vietnamese Baptist Church",
yearsShipping: 2,
languages: ["C#", "Python", "SQL", "TypeScript", "JavaScript", "C++"],
stack: [".NET 9", "ASP.NET Core", "Entity Framework Core", "FastAPI", "PyTorch", "Detectron2", "Vue 3", "React"],
focus: "Full-stack engineer modernizing .NET platforms used in 100+ countries",
career: [
{ at: "Vietnamese Baptist Church", role: "AI Software Engineer (Volunteer)", when: "Sep 2026 – Present" },
{ at: "OneHope", role: "Junior Software Engineer (Full Stack)", when: "Jan 2025 – Sep 2026" },
{ at: "OneHope", role: "Software Engineer Intern", when: "May 2024 – Aug 2024" },
],
education: "B.S. in Computer Science: Software Engineering, Liberty University",
};projects.aiFollicleDetector() Led a 6-person team delivering a cell-imaging tool for a university biology lab
- Gathered requirements from researchers and fine-tuned Mask R-CNN (Detectron2) to automate manual follicle counts used to judge treatments
- Scored the model against a held-out test set of 126 follicles, detecting 112 (F1 0.89)
- Prevented data leakage with a splitter keeping each of 29 animals in one split within 0.5 points of a 60/20/20 split target (50/17/17 images)
- Stack: Python, Detectron2, PyTorch, CVAT SDK, FastAPI, Docker
projects.vulnPriorityEngine() Turn a scanner's flat CVE dump into a risk-ranked, SLA-bucketed remediation queue
- Uses EPSS exploit-probability + CISA KEV + asset criticality to prioritize vulnerabilities
- Transforms unstructured scanner output into actionable remediation tasks
projects.advancedRagLlmApp() Self-contained implementation of an advanced AI application combining RAG, vector search, and LLM APIs
- Combines Retrieval-Augmented Generation (RAG), vector search, document ingestion, and prompt orchestration
- Stack: Python, CSS, HTML, JavaScript, Dockerfile, Makefile
projects.netcodeArena() Server-authoritative multiplayer netcode: a 2-player arena with client-side prediction
- Implements server reconciliation and snapshot interpolation, tested under injected latency and packet loss
- Engine-agnostic .NET core architecture
- Stack: C#, Docker, GitHub Actions, UDP networking, xUnit
projects.fieldOpsSqlCostAudit() Tested, layered analytical-SQL codebase computing cost-to-serve and technician utilization
- Computes cost-to-serve, technician utilization, and SLA-breach rates over messy field-ops data
- Surfaced in live Tableau dashboards
- Stack: Python, SQL, dbt, PostgreSQL, Tableau, Dockerfile
projects.contextLens() Engineers waste time re-deriving why code is written a certain way
- Surfaces tribal knowledge from PR comments and commit messages directly in the source line they explain
- Stack: TypeScript, Git log/blame parsing, GitHub REST API, SQLite, VS Code Extension API
Sep. 2026 – Present
Shipped a production PWA for 20+ users with live chat, push notifications, and RAG-based Voice AI answering Vietnamese questions from English PDFs with citations, isolating each tenant's data via Postgres RLS.
Built and deployed a resilient AI microservice with multi-provider failover and exponential backoff on Oracle Cloud using Docker, ensuring 99.9% uptime during LLM outages; cut pgvector search from 1.2s to 0.25s using connection pools.
Ran user acceptance interviews with congregation members, translated their feedback into a prioritized backlog, and documented expected behavior per feature, so every release shipped against criteria the users had approved beforehand.
Jan. 2025 – Sep. 2026
Shipped internal developer tooling across 14 .NET repos: a flag-gated request tracer (OpenTelemetry, Jaeger) making any slow request traceable from client to server to exact SQL; 3 teammates used it to cut load times 50–70% on 3 pages.
Improved the Planning dashboard load time 20x (10s to 0.5s) and cut its chart's database reads 93% by fixing a badly misplanned SQL Server query and removing duplicate API calls, verifying no result or permission changed.
Remediated a critical security vulnerability by optimizing O(n²) algorithms to O(n) and adding request throttling to prevent CPU exhaustion attacks, while delivering 40+ other production features and fixes across C# .NET services.
May 2024 – Aug. 2024
Executed a complex data migration recovery with a Python ETL pipeline on the Bynder API, restoring 150,000 asset relationships for 22,666 assets and saving research and marketing teams an estimated 200+ hours of manual re-linking.
Delivered 15+ production features and fixes by connecting data sources to an enterprise reporting engine, enforcing row-level security, debugging SQL generation errors, and writing automated unit tests in Azure DevOps CI/CD pipelines.
email · linkedin · website · github
- vuln-priority-engine — Turn a scanner's flat CVE dump into a risk-ranked, SLA-bucketed remediation queue using EPSS exploit-probability + CISA KEV + asset criticality.
- supply-chain-optimizer — AI-powered inventory management system that predicts demand 8 weeks ahead, optimizes stock levels, and reduces carrying costs by 40%.
- advanced-rag-llm-app — This project is a self-contained summer-internship-level implementation of an advanced AI application that combines Retrieval-Augmented Generation (RAG)…
- fraud-detection-system — Protecting $20B+ in annual e-commerce transactions with ML-powered fraud detection
- elt-pipeline-dbt-snowflake-airflow — A production-grade ELT pipeline that extracts data from multiple public APIs (Reddit, OpenWeather, and a SaaS PostgreSQL database), loads raw data into…
- netcode-arena — Server-authoritative multiplayer netcode: a 2-player arena with client-side prediction, server reconciliation, and snapshot interpolation, tested under…
- field-ops-sql-cost-audit — A tested, layered analytical-SQL codebase computing cost-to-serve, technician utilization, and SLA-breach rates over messy field-ops data, surfaced in live…
- contextlens — Engineers waste significant time re-deriving why code is written a certain way — the tribal knowledge lives in PR comments and commit messages disconnected…
- talentflow-recruitment — Smart Recruitment That Reduces Time-to-Hire by 50%
- flavormetrics-restaurant-analytics — Data-Driven Restaurant Management That Boosts Profits by 23%