Le Tran Minh Dat

AI-NATIVE FULL-STACK DEVELOPER

Ho Chi Minh City, Vietnam[email protected]GitHubLinkedIn

SUMMARY

Full-stack engineer with six production systems shipped in the past year, solo or leading small teams: a bank-wide LLM gateway, a VAS-compliant accounting ERP used daily by 20 accountants, an on-prem RAG advisor for a 19-branch academy, a real-time patient-monitoring vision system, a voice ordering agent and an AI trading platform delivered in 20 days. I own the full stack on every one: data model, backend, frontend, infrastructure and deployment.

EXPERIENCE

LLM Gateway Engineer

2026 – Present

Enterprise client (Banking) · Outsourced contract · 2-engineer team

Owned: architecture and the large majority of the custom layer, including its test suite and technical documentation.

  • Architected a bank-wide LLM gateway on LiteLLM as the single control point between every internal application and major LLM providers; isolated all bank-specific code in a separate 95-file extension layer with a single unlock hook into the core, so upstream releases merge cleanly.
  • Delivered enterprise access control: per-user, per-team and per-project budgets, SSO with the bank's identity provider, role-based permissions and an immutable audit log, plus bidirectional guardrails that mask or block phone numbers, national IDs, account numbers and cloud credentials.
  • Engineered cost governance: per-call spend reconciled against actual AWS invoices and tagged for department-level allocation; automatic model failover, budget-aware downgrade to cheaper models, and OAuth-based access to developer AI subscriptions.

AI Agent Consultant Engineer

2025 – Present

Hướng Nghiệp Á Âu · Freelance · Solo, end-to-end

Owned: everything — backend, frontend, prompt pipeline, data model, infrastructure and deployment.

  • Built a 24/7 RAG enrollment advisor for a 19-branch academy (100,000+ learner base), handling 202 leads and ~1,800 messages to date, with an advisor cockpit for human-in-the-loop review.
  • Hybrid retrieval fuses local ONNX embeddings (CPU, zero API cost) with Vietnamese full-text search; a CRM-sourced allow-list and topic pinning prevent hallucinated courses.
  • Instrumented every LLM call (tokens, cost, latency, errors) and migrated providers on measured A/B evidence.
  • Deployed on-prem inside the client's VPN-only LAN under systemd, integrated with the internal CRM, hardened with auto-restart and a 9-suite E2E test gate.

Accounting Application Developer

2025 – Present

Enterprise client (Education) · Freelance · 2-engineer team

Owned: full backend, frontend and database design, delivered jointly with one other engineer.

  • Built a VAS-compliant ERP (.NET 10, PostgreSQL, React) in daily production use by 20 accountants; authored the SRS and 164 regulatory test cases it is audited against (audit passed, in trial period).
  • Shipped double-entry ledger logic and a full audit trail through Docker + AWS EC2 + GitHub Actions CI/CD, ahead of the 1 Jan 2026 regulatory deadline.

AI Team Lead, Computer Vision

Jun 2025 – Jan 2026

Whammy Tech · Contract · Led 4 engineers

Owned: all AI work (detection models, inference pipeline, LLM reporting) and technical documentation; advised the backend, frontend and mobile engineers.

  • Led all AI development on IPBMS, a real-time fall/seizure detector on live RTSP streams (YOLO pose estimation plus LLM incident reports); designed a 5-frame temporal validation strategy to suppress single-frame false alarms.
  • Deployed and operated on-premise, live on 2 production cameras.

AI Agent Engineer

2025 – Present

Ăng Ê Coffee · Contract · Solo, self-hosted

Owned: everything, from agent logic to the self-hosted server it still runs on.

  • Built a real-time voice ordering agent (LiveKit, TTS) in daily production use, with hand-built multi-LLM failover and a self-learning RAG loop.
  • Rolled out in phases (text → voice → real-time voice); self-hosted and still operates the stack.

Freelance Software Engineer, Team Lead

Oct 2025

NEURADAO · Led 3 engineers

Owned: AI trading logic, AI chatbot and trading-automation backend; one backend and one frontend engineer on the rest.

  • Led a 3-engineer team from zero to a production AI trading platform in 20 days, setting the architecture and delivery plan and owning AWS deployment (PostgreSQL) through go-live.

Platform Engineer (personal infrastructure)

2026 – Present

Cicca Fleet · Self-hosted k3s homelab

Owned: design and operation of a multi-site Kubernetes cluster that hosts my public sites, including this one.

  • k3s across cloud VPS nodes and an ARM64 Orange Pi at home, joined over a Tailscale mesh with ACLs separating personal and client machines.
  • Fully GitOps: an ArgoCD app-of-apps syncs every workload from a Git repo; in-cluster container registry backed by MinIO; NFS StorageClass on a Synology NAS.
  • Observability with Grafana, Mimir, Loki and Alloy (node, Docker, GPU and NAS/SNMP metrics) and alert rules managed as code.
  • Public traffic enters through Cloudflare Tunnel — no inbound ports opened on the nodes.

IT Support Technician

Jan 2022 – Dec 2024

University IT · Part-time

Supported university-wide IT operations across Academic Affairs, Examinations, HR, Accounting and Administration for cohorts K15–K20: hardware, networks, classroom and exam systems, user accounts, and onboarding new student IT support members.

SELECTED PROJECTS

  • Bank-wide LLM Gateway · BANK-WIDE CONTROL POINT

    Single control point between every internal bank application and major LLM providers, built on LiteLLM. All bank-specific code lives in a separate 95-file extension layer behind one hook, so upstream releases merge cleanly. Per-user/team/project budgets, SSO, RBAC, immutable audit log, bidirectional PII guardrails, AWS-invoice cost reconciliation and budget-aware model failover.

    LiteLLM · SSO / RBAC · Guardrails · AWS Cost Governance

  • On-prem RAG Enrollment Advisor · 202 LEADS · ~1,800 MESSAGES

    24/7 advisor for a 19-branch academy (100,000+ learner base): 202 leads and ~1,800 messages so far, with an advisor cockpit for human-in-the-loop review. Hybrid retrieval fuses local ONNX embeddings with Vietnamese full-text search; a CRM allow-list prevents hallucinated courses.

    Hybrid RAG · ONNX Embeddings · Human-in-the-loop · systemd / VPN

  • VAS-compliant Accounting ERP · 20 ACCOUNTANTS DAILY

    Double-entry accounting ERP (.NET 10, PostgreSQL, React) in daily use by 20 accountants. Authored the SRS and 164 regulatory test cases it is audited against (audit passed); shipped through Docker, AWS EC2 and GitHub Actions ahead of the 1 Jan 2026 deadline.

    .NET 10 · PostgreSQL · React + TanStack · GitHub Actions

  • IPBMS — Real-time Patient Monitoring · 5-FRAME TEMPORAL VALIDATION

    Real-time fall/seizure detection on live RTSP streams using YOLO pose estimation plus LLM incident reports. Designed a 5-frame temporal validation strategy to suppress single-frame false alarms; deployed on-premise, live on 2 production cameras.

    YOLO Pose · OpenCV · RTSP · LLM Reports

  • Real-time Voice Ordering Agent · DAILY PRODUCTION USE

    Voice ordering agent for Ăng Ê Coffee in daily production use (LiveKit, TTS) with hand-built multi-LLM failover and a self-learning RAG loop. Rolled out text → voice → real-time voice; self-hosted and still operated by me.

    LiveKit · TTS · Multi-LLM Failover · Self-learning RAG

  • NEURADAO — AI Trading Platform · SHIPPED IN 20 DAYS

    Led a 3-engineer team from zero to a production AI trading platform in 20 days: architecture, delivery plan, AI trading logic, chatbot and trading-automation backend, deployed on AWS with PostgreSQL through go-live.

    AI Trading · Chatbot · AWS · PostgreSQL

  • Cicca Fleet — Multi-site k3s Cluster · 4 NODES · MULTI-ARCH · GITOPS

    Personal Kubernetes fleet spanning cloud VPS nodes and an ARM64 Orange Pi at home, meshed over Tailscale and managed fully by GitOps (ArgoCD app-of-apps). In-cluster container registry on MinIO, NFS storage from a Synology NAS, Grafana + Mimir + Loki + Alloy observability with alerting, public apps via Cloudflare Tunnel — this website is served from it.

    k3s · ArgoCD · Tailscale · Grafana LGTM

SKILLS

LLM / AI
Multi-provider LLM routing & fallback · Hybrid RAG · Guardrails & PII masking · Human-in-the-loop agents · YOLO / OpenCV / RTSP · Prompt & context engineering
FULL-STACK
TypeScript · Python · C# / .NET 10 · NestJS / Node.js · React + TanStack · PostgreSQL · pgvector · Redis · REST API design
CLOUD & OPS
Kubernetes (k3s) + ArgoCD · Docker · AWS EC2 · GitHub Actions CI/CD · Observability · On-prem / VPN deployment · RBAC & audit logging · E2E test gates
SPOKEN
Vietnamese (native) · English (professional)

EDUCATION & AWARDS

  • Top 4 — Viettel AI Race AWARD

    Hackathon organised by Viettel Group.

  • B.Eng. Software Engineering — FPT University 2026

    Graduated 2026 from FPT University, Ho Chi Minh City · GPA 3.47 / 4.0.

  • Regulatory audit passed 2026

    VAS-compliant ERP audited against the 164 regulatory test cases I authored; in daily use by 20 accountants.

  • Production platform in 20 days 2025

    Led a 3-engineer team from zero to a live AI trading platform on AWS for NEURADAO.

  • Leader — IT Support Technician 2022

    Worked as a part-time IT Support member while studying, supporting university IT operations across multiple departments.