Senior AI Engineer | Senior Software Engineer
π Hai BΓ TrΖ°ng, Hanoi, Vietnam Β β’Β πΌ 7+ Years Experience Β β’Β π€ Enterprise AI Agent Mesh Β· A2A Β· MCP
Senior AI Engineer & R&D Solution Architect with 7+ years designing, architecting, and building high-performance, fault-tolerant backend systems, with deep expertise in microservices architecture using Java, Golang, and Python. Proven track record as an R&D-focused architect and technical lead, pioneering the design of highly complex, distributed core platforms β including Enterprise AI Agent Mesh systems. I specialize in researching and establishing new technical standards for Agent-to-Agent (A2A) protocols and Model Context Protocol (MCP) integrations β solving unprecedented architectural challenges, pushing the boundaries of LLM capabilities, and architecting fault-tolerant, scalable multi-agent ecosystems from the ground up.
Top skills: Artificial Intelligence (AI) Β· Solution Architecture Β· Agentic AI Development Β· Technical Solution Design Β· System Design
| Category | Technologies & Tools |
|---|---|
| Languages |       |
| AI / Agentic |     A2A AG-UI LLM & Prompt Engineering |
| AI Tooling |    Skills / Agents / Hooks uv Jinja |
| Backend Frameworks |     |
| Frontend |      |
| Mobile |  TornadoFX |
| Data & Streaming |      TiDB Dremio SSE |
| Cloud & DevOps |   -326CE5?style=flat-square&logo=kubernetes&logoColor=white){kind=icon}  OAuth / MSAL |
| Observability |    Langfuse Loguru |
| Architecture | AI Agent Mesh Multi-Agent Systems Event-Driven (EDA) Microservices Rule Engines Hybrid Service Registry Multitenancy Circuit Breakers |
Jun 2026 β Present | TechnoPark Tower, Hanoi, Vietnam
The Vingroup company delivering unified technology & AI solutions across the entire corporation. Researching and developing the group-wide shared AI platform β defining foundational AI standards, agent-mesh infrastructure, MCP tooling, and reusable AI capabilities adopted across the group's businesses.
Tech:
PythonGoogle ADKMCPGemini / Vertex AIFastAPIClaude Code
Sep 2020 β Jun 2026 | Hanoi, Vietnam
Evolved from Senior Software Engineer into Solution Architect & R&D Lead, pioneering distributed platforms β Enterprise AI Agent Mesh and high-performance microservices.
πΈοΈ Agent Space β Enterprise AI-Agentic Platform (Agent Mesh) (Jul 2025 β Jun 2026)
- Spearheaded a 9-month R&D initiative to architect and deploy a proprietary Python-based 8-Layer Enterprise Agent Mesh Platform, using a customized Agent Development Kit (ADK) to orchestrate hundreds of AI agents as collaborative virtual employees across Root / Parent / Sub-agent tiers.
- Architected a Hybrid Service Registry (two-layer pattern: durable Firestore Profile Layer + ephemeral Redis Pulse Layer with TTL liveness), achieving <100ms discovery latency and 5 load-balancing strategies (Random, Round-Robin, Least-Conn, Weighted, Consistent Hash).
- Designed dynamic tool integration via Model Context Protocol (MCP) for seamless access to external APIs, knowledge bases, and code-lookup services; migrated dependency/venv management to uv to accelerate the R&D cycle.
- Engineered a rigorous 4-Layer Loop Prevention system (app-level hop limits, network circuit breakers, trace-hash deduplication, real-time anomaly detection).
- Extended the ADK with long-term memory management & compaction (token budgeting + waterfall thresholds) to prevent LLM context exhaustion and dramatically cut API costs.
- Integrated an observability stack β Langfuse v3 (LLM cost/token), OpenTelemetry (context propagation), Loguru (distributed JSON tracing).
- Deployed a unified multi-adapter Gateway (Teams MSAL OAuth, Google Chat ADC, Web UI) with AG-UI / A2UI dynamic rendering over SSE β >10K events/sec throughput, <16ms UI render.
- Shipped 2 production use cases: an Internal Operations Virtual Assistant and a Delivery Assistant integrated with the Transport Management System (TMS) for autonomous scheduling with Human-in-the-Loop (HITL).
- Tech Stack:
PythonGoogle ADKMCPA2AAG-UIFirestoreRedisSSELangfuseOpenTelemetry
π§ͺ AISDLC β AI Integration in Enterprise SDLC (Jan 2025 β Jun 2026)
- Partnered directly with a 20-member executive board (C-level + senior domain experts) to formulate, evaluate, and execute AI integration strategy for the enterprise Software Development Life Cycle.
- Benchmarked cutting-edge LLMs (Gemini 3 Pro/Flash, Vertex AI) and self-hosted models (Gemma, GLM, Qwen) for performance, quality, and ROI.
- Engineered specialized AI agents (OM Git Agent, OM Jira Agent, AI Manual Tester) with custom personas, cognitive methods, and skill sets.
- Implemented robust AI guardrails using dynamic LLM-based command-safety evaluation (replacing static regex) to mitigate prompt injection; enforced a strict read-only execution protocol β zero unauthorized code modifications, outputs limited to actionable docs and review reports.
- Built end-to-end automated workflows: agents autonomously perform code reviews, audit commit history, verify Jira compliance, and generate proactive security/compliance reports.
- Tech Stack:
Gemini 3Vertex AISelf-hosted LLMsMCPAI Guardrails
π Metric Platforms β Centralized Metric Engine (Sep 2024 β Jun 2025)
- Led and mentored a cross-functional team of 12 engineers to overhaul and modernize the core Metric Platform β a centralized hub for KPI and promotional-data calculations.
- Engineered high-throughput dynamic ingestion pipelines across REST APIs, Data Lakes (Dremio), Oracle, TiDB, and Kafka.
- Merged the metric core with a custom Rule Engine + drag-and-drop Rule Builder UI, empowering Ops to configure complex KPI formulas without engineering intervention.
- Optimized multitenancy for strict data isolation and fault tolerance; added execution resilience with Temporal retries/state management and Splunk + Google Chat alerting.
- Tech Stack:
GolangTemporalKafkaDremioOracleTiDBSplunk
ποΈ HRM β Leave & Attendance Management System (Sep 2023 β Sep 2024)
- Architected a highly available, real-time multi-service ecosystem (Golang + Temporal) automating attendance/timesheet calculation for a large distributed field-sales workforce (SA, KA, CTV, Telesales).
- Designed the core of a proprietary Rule Engine enabling HR to configure legally-compliant attendance/OT/leave policies β cutting manual processing time by 80%.
- Led design of bidirectional data sync between the microservices ecosystem and the legacy SAP ERP.
- Tuned GCloud (Docker/Kubernetes) to handle 1000+ concurrent check-in/out requests at peak.
- Tech Stack:
GolangTemporalGKEDockerSAP ERP
π Sale Tools (B2B2C) & VinID Back Office (Oct 2020 β Aug 2023)
- Designed and aggressively optimized critical backend microservices for a large-scale B2B2C retail/consumer platform (Sale Tools & VinID Back Office) using Java (Spring Boot) and Golang.
- Containerized legacy monolith components; managed deployments with Docker + Kubernetes on GCP.
- Refactored data-access patterns and indexing strategies, reducing critical API response times by 75%.
- Collaborated within a 25β30 developer agile division with PM, QA, and BA to ship user-centric features for the Vin Group digital ecosystem.
- Tech Stack:
Java (Spring Boot)GolangDockerKubernetesGCP
Aug 2019 β Sep 2020 | Hanoi, Vietnam
- Executed the migration and modernization of critical financial modules from legacy COBOL to Java, ensuring absolute data integrity and 99% uptime.
- Tech Stack:
JavaCOBOL
Aug 2018 β Jul 2019 | Hanoi, Vietnam
- Designed an intelligent threat-identification engine using custom data pipelines and highly optimized MongoDB / Elasticsearch clusters.
- Tech Stack:
MongoDBElasticsearch
Mar 2018 β Aug 2018 | Hanoi, Vietnam
- Began my engineering career building and maintaining enterprise web applications in Java.
| Project | Description | Stack |
|---|---|---|
| product-spec Β Β·Β live demo β | AI harness (skills + agents + hooks) for Claude Code that turns partner requirements into a structured, frontmatter-linked, fully traceable product-spec hierarchy β closed loop: Define β Validate β Critique β Update. | Python Claude Code |
| human-analyzer Β Β·Β live demo β | Clinical-grade Character Profile Intelligence System β 6 event-driven modules (evidence scoring T1βT5, psychological case-formulation, growth analysis, content generation) with PII/claim safety gates & 700+ tests. | Python Event-Driven Knowledge Graph |
| sdlc-harness Β Β·Β live demo β | A file-based SDLC discipline for Claude Code: prose skills/rules guide the agent, while fail-closed PreToolUse hooks gate every push/PR/ship on real artifacts β plan β cook β test β ship with reviewerβ author approval, enforced redβgreen TDD (797 tests + invariants), and an append-only audit trail with actor attribution. 14 hooks Β· 31 scripts Β· ~46 hs:* skills Β· 14 agents. Interactive bilingual (EN/VI) showcase. |
Python Claude Code MCP Hooks |
| Agent Space π | Enterprise 8-layer AI Agent Mesh orchestrating hundreds of distributed agents β Google ADK + MCP with a hybrid Firestore/Redis service registry. (proprietary / enterprise) | Python Google ADK MCP FastAPI |
- π Potential Leader Nomination β One Mount Group, Dec 2025 β for exceptional technical leadership and consistent delivery (two consecutive years).
- π Top Performance Ratings β top-tier appraisals (A / A+) for three consecutive years (2023, 2024, 2025).
- β‘ Accelerated Engineer's Degree β completed the 5-year engineering curriculum in 4 years, graduating 1 year ahead of schedule.
-005BAC?style=flat-square&logo=ets&logoColor=white){kind=icon}
-E60012?style=flat-square&logo=japanpost&logoColor=white){kind=icon}
- π± Mentor β VietSeeds Foundation (2019 β Present)
- π©Έ Mentor β Hanoi Association of Youth Blood Donor Recruiters (HAYBDR) (2015 β Present)
π Engineer's Degree in Computer Software Engineering β Grade: Very Good Hanoi University of Science and Technology (HUST) β 2015 β 2019
Built strong foundations in algorithmic problem-solving, software design patterns, and systems engineering. Completed the 5-year engineering curriculum in 4 years. Activities: VietSeeds Foundation Β· Hanoi Association of Youth Blood Donor Recruiters Β· Hoa Trang Nguyen Club.
Let's build something extraordinary in AI. Get in touch via LinkedIn or Email!