Portrait of Hoang Nhat Duy Le

M.S. Computer Science · University of Cincinnati · CincySI Lab

I build scalable, production systems.

Software engineer with 4+ years in production, from manufacturing software serving 30+ customers to a machine-learning rollout across First Solar sites in the US, India, and Southeast Asia. Now an M.S. Computer Science student and researcher at CincySI Lab, University of Cincinnati, working on multi-agent LLM systems.

  • Research opportunities, Summer 2027
  • Full-time roles after graduating May 2028
Résumé (PDF)
Years building software
4+
Industry employers
4
Customers served
30+
Papers under review
2

I work on multi-agent LLM systems at the CincySI Lab (University of Cincinnati): how to audit what agents claim, and how their confidence behaves as messages pass between them.

  • Under review · NAACL 2027

    Auditing whether the evidence multi-agent LLM systems cite is what actually drove their decisions.

    Title withheld while under review.

  • Under review · NAACL 2027

    How confidence changes as claims pass between agents in multi-agent LLM pipelines.

    Title withheld while under review.

CincySI Lab

Graduate Researcher

Oct 2026 – Present

Achievements

  • Researching multi-agent LLM systems with Prof. Justin Zhan in the Super Intelligence Lab at the University of Cincinnati.
  • Two manuscripts on auditing and calibrating multi-agent LLM pipelines are currently under review.
  • Multi-Agent Systems
  • LLMs
  • Evaluation
  • Python
  • Daily

    What I reach for first

    • RAG
    • Agent orchestration
    • LLM evals
    • Agent security & provenance
    • Python
    • FastAPI
    • PostgreSQL
    • TypeScript
    • React / Next.js
    • Docker
  • Working knowledge

    Comfortable, not daily

    • Kubernetes
    • Redis
    • CI/CD
    • Vector stores
    • Azure
    • C# / .NET
    • Angular
  • Earlier career

    Where I started

    • MES & PLC integration
    • PHP
    • JavaScript
    • SQL Server
    • C++
    • Java
AEGIS pixel-art illustration: an "ignore rules" attack is blocked by the firewall while a student learns why.

★ Winner · Best Use of Auth0 · MakeUC 2026Nov 2026

AEGIS: an agent firewall that teaches

A rule-based firewall that intercepts every AI agent tool call, blocking injected commands and secret leaks before they run and logging the rule behind each decision. A classroom on top lets students attack level chatbots, then take a quiz generated from their own session.

My role: I built Auth0 sign-in with three server-enforced roles (student, teacher, firewall admin) and the teacher/student dashboard.

  • Next.js
  • Python
  • PostgreSQL
  • Auth0
  • Docker
  • University of Cincinnati

    Cincinnati, Ohio, United States

    • Master of Science in Computer Science
    • Research advisor: Professor Justin Zhan
    • Research focus: multi-agent LLM systems at CincySI Lab

    Aug 2026 – May 2028 (expected)

  • University of Toledo

    Toledo, Ohio, United States

    • B.S. Computer Science & Engineering · Minor in Data Science
    • GPA 3.89
    • Completed 5 co-ops (full- and part-time)

    Aug 2018 – May 2023

  • Hoang is a reliable, high-output contributor who consistently turns complex work into tangible results. His technical work has driven real value for the team, and his attention to detail has helped bring new internal tools online smoothly. He's also shown strong agility managing competing priorities, delivering complex data integration work in short timeframes without sacrificing quality.

    Aaron BollingerSoftware Engineer III - MES, First Solar
  • One of the many positive traits that Le has, is his ability to take what we have as a vision for a project and turn it into reality. He created a few dashboards that were able to allow us to better understand where our business stands with not only target dates, but also easily view planned to actual data side by side. He did this, while creating convenient and creative ways to input data easily.

    Heysam SalamiDirector of Software Operations, Nysus Solutions
  • Hoang demonstrated the ability to quickly adapt and pick up on our coding practices. He has been exceeding expectations on given tasks and has shown a strong interest in learning, through his clarifying questions and take time to work through his own problems. He has been an enjoyable team member and assimilated quickly into our culture.

    Barry FlemingSoftware Development Manager, Equity Trust Company

Ask my AI

Waking up the assistant. This can take about 30 seconds after inactivity.

AI-generated answers; please verify anything important.