Truong Le Vinh Phuc's Resume

Truong Le Vinh Phuc

AI Product Engineer • Product, AI/LLM & Cloud Platform

Ho Chi Minh city

Truong Le Vinh Phuc's profile picture

About

Technical Product Manager and Product Engineer who owns products end-to-end — from PRDs and roadmap to production code. Currently leads the Katalon Support Assistant as PO + engineering lead, where a multi-agent LLM system deflects 30% of customer cases without human escalation. Background spans AI/LLM products, cloud-native microservices, and blockchain platforms.

Work Experience

KATALON

2025 - Present

AI Product Engineer

  • Product Owner and Engineering Lead for Kally and the Katalon Support Assistant; own roadmap, architecture, delivery, and hands-on execution
  • Led Kally, a Slack-native ambient AI teammate that helps product and support engineers triage cases, analyze logs, create Jira tickets, draft Confluence articles, review GitHub pull requests, and monitor CI workflows
  • Designed the core security model where AI acts through each user's own authorized credentials instead of a shared service account, with approval-gated write actions, encrypted credential storage, access policies, and audit logs
  • Built multi-agent workflows powered by OpenCode for runbook execution, ticket creation, knowledge-base drafting, log analysis, GitHub PR review, CI analysis, and autonomous fix attempts
  • Architected a secure MCP proxy layer connecting Salesforce, Jira, Confluence, Google Drive, Slack, GitHub, Grafana, Metabase, PostHog, and LaunchDarkly while enforcing policy checks before upstream systems are touched
  • Delivered a cloud-native platform across 9 Google Cloud Run services, including a multi-container runner with an OpenCode sidecar and isolated Daytona sandboxes for untrusted command execution
  • Maintained an 11-package TypeScript monorepo with strict typing, shared schemas, 733 automated tests across 45 test files, and CI/CD through GitHub Actions, Cloud Build, Artifact Registry, and Workload Identity Federation
  • Product Engineer and Support Lead for TrueTest, Katalon's autonomous testing platform that turns production web traffic into executable end-to-end test suites — ran customer-facing communication and issue triage, then closed the loop by implementing the fixes and shipping the features myself
  • Shipped full-stack into TrueTest's production stack — Java/Quarkus backend services and the internal React operations console — including session search across both the Athena and ClickHouse execution paths with server-side query validation, a raw-data inspection view, and global AUT search with per-token matching and active-first ranking, reused on the Support Mode console
  • Exposed Smart text exclusions in TrueTest Mode with a newline-delimited selector input, so CSS selectors containing commas and spaces round-trip without silent corruption
  • Built the diagnosis tooling support and operations use to triage live customer sessions, cutting time-to-root-cause on escalations and removing the back-and-forth that used to block every investigation
  • Held a test-first bar on a large multi-team codebase, pairing every TrueTest change with unit coverage across React components and Java query builders so support-critical paths could not silently regress
  • Partnered with Product, Engineering, and Account Management to surface customer insights, prioritize improvements, and expand internal/self-service support knowledge
  • On-site/Remote
  • AI/LLM
  • Product Owner
  • Full-stack
  • DevOps
  • Support Lead

FPT SOFTWARE

2024 - 2025

Software Engineer Intern

  • Supported CI/CD improvements and streamlined deployment workflows for a Spring Boot API system
  • Optimized PostgreSQL queries and improved system efficiency through performance-oriented fixes
  • Worked in Agile/Scrum with Jira, contributing to sprint planning and delivery
  • On-site
  • Java
  • Spring Boot
  • PostgreSQL
  • CI/CD

Education

WorldQuant University

2026 - Present
Graduate Certificate in Applied AI Fundamentals (AAIF)

University of Information Technology - Vietnam National University, Ho Chi Minh city

2022 - 2026
Bachelor of Science in Software Engineering • GPA 3.8/4

Certificates

Skills

  • PRD Writing
  • User Stories & Acceptance Criteria
  • Backlog Prioritization
  • Roadmap Planning
  • Release Planning
  • Agile/Scrum
  • Stakeholder Communication
  • Jira
  • Confluence
  • Notion
  • .NET
  • Java
  • Node.js
  • Go
  • Python
  • TypeScript
  • React
  • Next.js
  • Tailwind
  • REST
  • gRPC
  • Microservices
  • Event-driven Systems
  • System Design
  • Databricks SQL
  • OAuth2/JWT
  • AWS (EKS/ECS/CloudFormation)
  • Google Cloud Run
  • Docker
  • Kubernetes
  • Helm
  • Istio
  • Terraform
  • Terragrunt/Atlantis
  • GitHub Actions
  • Jenkins
  • ArgoCD
  • SOPS/Secrets Manager
  • LangGraph
  • LangChain
  • Anthropic Claude
  • Multi-agent Systems
  • MCP
  • RAG
  • Prompt Engineering
  • LLM Observability & Evals

Side projects

TrueTest – Autonomous E2E Testing from Production Traffic

Support Lead and full-stack engineer on a platform that turns recorded production web traffic into executable test suites. Built session search across both AWS Athena and ClickHouse execution paths with server-side query validation, and shipped the session-ID search and raw-data inspection views support uses to triage live customer sessions — cutting time-to-root-cause. Added global AUT search (per-token matching, active-first ranking) and exposed Smart text exclusions in TrueTest Mode, shipping small AI-assisted PRs each paired with unit tests.

  • Java
  • Quarkus
  • React
  • AWS Athena
  • ClickHouse
  • SQL
  • Jest
  • Docker

Scriptorium – AI Teammate That Asks Before It Acts

Open-source AI teammate for Slack, Jira, Confluence, and GitHub where rules are enforced in code, not by prompt: one policy check per tool call, signed single-use approvals bound by hash to the exact arguments, exactly-once writes, a no-citation-no-claim grounding contract, and a hash-chained audit log. Built AI-native — an agent loop ships up to five changes every 30 minutes (research, build with evals, adversarial review, mutation-tested guards, then typecheck, 500+ evals, and CI) before a person merges each PR. 253 commits, 8 packages, 11 skills.

  • TypeScript
  • Node.js
  • Slack
  • Jira
  • Confluence
  • GitHub
  • MCP
  • Anthropic Claude

Katalon Support Assistant

Owned product vision and shipped a customer-facing AI support platform end-to-end as PO and engineering lead. Three specialized agents (MCP, Ticket, Research) route queries across docs, live web search, and Salesforce CRM, deflecting 30% of cases without human escalation. Provider-agnostic LLM layer (Anthropic API or AWS Bedrock, per tool), deployed on AWS ECS via CloudFormation and ArgoCD GitOps with JWT-derived agent identity and CSAT feedback loops.

  • Python
  • FastAPI
  • LangGraph
  • Anthropic Claude
  • AWS Bedrock
  • MCP
  • Next.js
  • ArgoCD

Kally – Ambient AI Teammate in Slack

Slack-native AI teammate that triages cases, analyzes logs, drafts Jira tickets and Confluence articles, reviews GitHub PRs in real git worktrees, and fixes failing CI — acting through each user's own credentials, never a shared service account. Approval-gated writes re-authenticate as the requester, backed by an AES-256-GCM credential vault and an append-only audit log. 11-package TypeScript monorepo, 733 tests, 9 Cloud Run services; built downstream of scoutqa-dot-ai/thor.

  • TypeScript
  • OpenCode
  • MCP
  • Google Cloud Run
  • Slack
  • Daytona
  • AES-256-GCM

Katalon Academy – Learning & Certification Platform

Engineering lead owning the production path of a platform serving 100,000+ learners across 110+ courses. Built the certification scoring and data-sync backend for two new True Platform certifications, generalising the awarding engine to N-of-M requirement groups with a Databricks connector, and proved four live certifications byte-identical across 2,424 values before shipping. Secrets flow SOPS → AWS Secrets Manager → Secrets Store CSI, never into git.

  • PHP
  • WordPress
  • Svelte
  • Databricks
  • Terragrunt
  • Kubernetes
  • ArgoCD
  • SOPS

Customer Service Dashboard

Delivered production infrastructure for an internal SSO-gated support command center unifying Salesforce case triage, Databricks-backed CSAT/response-time KPIs, and Jira-linked product feedback. Authored Terraform managing ~35 AWS resources (ECS Fargate, ALB, ACM/Route 53, DynamoDB, Amplify) with 7 CloudWatch alarms and SNS alerting, across 55 merged PRs.

  • FastAPI
  • Next.js
  • Terraform
  • AWS ECS Fargate
  • DynamoDB
  • Salesforce
  • Databricks
  • GitHub Actions

llmfit – Web Dashboard for Hardware-Aware LLM Recommendation

Designed and shipped the first-ever web dashboard for llmfit, a 25K+ star Rust project that recommends LLM models based on a user's hardware. Introduced a new `llmfit serve` HTTP entrypoint exposing system specs and ranked recommendations over REST, then built the dashboard with side-by-side model comparison, advanced filtering, multi-theme support, and Ollama integration.

  • React
  • JavaScript
  • Rust
  • REST APIs
  • Open Source

Cloud-Native E-Commerce Platform

Architected a 5-app microfrontend (Module Federation + Nx) and 4 .NET 10 microservices behind an Ocelot gateway on Clean Architecture + CQRS, event-driven via RabbitMQ. Provisioned AWS EKS with modular Terraform and IRSA for zero static credentials; single Kustomize tree validated in CI, Helm charts, 10 GitHub Actions workflows with Trivy/CodeQL and SBOM, and Istio mTLS with full observability. RFC 7807 errors, unit/integration/k6 load tests, and an Astro docs site. Reduced deploy time to <30 min and infra cost by 60%.

  • AWS EKS
  • Terraform
  • Kubernetes
  • Helm
  • Istio
  • .NET 10
  • Module Federation
  • Prometheus

GreenLedger – Carbon Credit Management Platform

Designed and shipped a 6-service Go microservices platform for carbon footprint tracking and credit management. Implemented JWT + RBAC, Kafka event-driven communication, and per-service PostgreSQL with Redis caching, on Kubernetes with a Traefik gateway, auto-scaling, Prometheus/Grafana observability, and security-focused CI/CD.

  • Go
  • Gin
  • gRPC
  • PostgreSQL
  • Redis
  • Kafka
  • Kubernetes
  • Traefik

Machine Learning Experiment Management

Built an ML experiment management system for MNIST classification with real-time training monitoring. Designed a web interface for hyperparameter configuration and used WebSocket streaming for live training progress and metrics visualization.

  • Flask
  • Next.js
  • React.js
  • OpenCV
  • TensorFlow
  • WebSocket
  • MySQL

Blockchain-based Online Notarization Platform

Led backend development for a B2B SaaS e-notarization platform with NFT-based tamper-proof records, Gemini-powered document verification, e-signatures, and real-time video sessions. Built two blockchain implementations — Ethereum (Solidity via Alchemy), then re-architected to Solana with Anchor and Bundlr/IPFS. Managed a 10-developer team — won First Prize at IT Hackathon 2025.

  • React.js
  • Node.js
  • React Native
  • MongoDB
  • Ethereum
  • Solidity
  • Solana
  • Anchor

Emotion Detection Web App using AI

Built a real-time emotion classification system detecting 5 emotions from portrait images. Implemented HOG feature extraction and CNN model training, exposed via a Flask API, with a cross-platform mobile UI in React Native.

  • Flask
  • React Native
  • OpenCV
  • TensorFlow

Awards

First Prize at IT Hackathon 2025 – Solana Pragmatic — Youth Union – Student Association, Faculty of Computer Science and Engineering, International University – VNUHCM
May 2025
Top 30 at Web3 HackFest 2025 – Web3 & AI Convergence — VBI Academy
Mar 2025
Consolation Prize at SEAPP Contest 2024 — Department of Software Engineering – UIT, VNUHCM
Nov 2024
2nd Runner-up at GDSC IDEA CONTEST 2023: THiNK — Google Developer Student Club – HCMUT
Jun 2023