ACB đang tuyển dụng HO - AI Engineer Expert — vị trí làm việc tại Hội sở (Hồ Chí Minh).
Thông tin tuyển dụng ACB
| Thông tin | Chi tiết |
|---|---|
| Vị trí | HO - AI Engineer Expert |
| Ngân hàng | ACB |
| Đơn vị | Phân tích Dữ liệu |
| Địa điểm | Hội sở (Hồ Chí Minh) |
| Mức lương | Thương lượng |
| Hạn nộp hồ sơ | 2026-07-31 |
Mô tả công việc
Key Responsibilities
Architecture & Technical Design
-
Own end-to-end architecture for the team‘s RAG, Agentic and Multi-Agent systems - retrieval strategy, agent orchestration patterns (Planner–Executor, Router, Verifier, Reviewer), tool-calling layer and state management.
-
Make and defend technology choices (vector DB, embedding model, orchestration framework, observability stack) with explicit trade-off analysis on capability, cost, latency and compliance.
-
Set engineering standards (coding conventions, inter-agent contracts, MCP tool schemas, minimum observability bar) and write ADRs for high-impact decisions.
-
Design reusable components (retrieval templates, MCP server skeletons, guardrail middleware, evaluation harnesses) so engineers don‘t rebuild foundations for each product.
Hands-on Engineering
-
Personally write code for the hardest components: agent orchestration core, tool-calling middleware, complex retrieval, guardrail engine and tricky integrations with bank legacy systems.
-
Prototype new architectural ideas and own the hard production incidents the team can‘t crack — hallucination edge cases, retrieval regressions, vLLM latency spikes, schema-breaking tool-call failures.
-
Stay hands-on: expect 40–50% of your time on direct coding. This is not a technical PM seat.
Mentorship & Code Review
-
Run high-quality code reviews that teach failure modes, cost awareness, observability and maintainability — not just check logic.
-
Mentor engineers 1-on-1 on technical growth and design decisions.
Stakeholder Representation, R&D and Reliability
-
Be the team‘s technical face with Risk, Compliance, IT Infrastructure and Business - translating regulatory and business constraints into design decisions, and technical trade-offs back to stakeholders.
-
Drive technical R&D: evaluate emerging techniques, run benchmarks, separate hype from value, and contribute to the division‘s technical roadmap.
-
Own production quality KPIs (hallucination rate, retrieval recall@k, tool success rate, latency, uptime); maintain evaluation frameworks that gate deployments; lead incident response and post-mortems.
-
Enforce banking-grade non-functional requirements: auditability, explainability and end-to-end traceability for internal audit and legal.
Core Requirements
Must-Have
-
Bachelor‘s degree or higher in Computer Science, AI, Data Science or a related field.
-
At least 10 years of professional software engineering, with at least 3 years hands-on production work on LLM systems / RAG / Agentic AI. Shipped at least 2 AI systems to production with real users - not POCs or demos.
-
Production-grade Python: async-first (asyncio, aiohttp), Pydantic data modeling, clean modular design.
-
Real RAG / Agentic experience: debugged retrieval quality issues for real, understands cosine vs. dot product.
-
LangChain / LangGraph: fluent with the state-machine model, custom nodes and practical edge cases.
-
Vector databases & LLM serving: non-trivial production pipelines on Qdrant, Milvus or Pinecone (index schema, namespaces, metadata filtering); hands-on deployment of open-source LLMs via vLLM with concurrency, batching and quantization (AWQ/GPTQ) tuned to cost-vs-latency targets.
-
MCP & containerization: designed or implemented MCP servers / tool schemas for agent consumption in production; comfortable with Docker, Compose and Kubernetes basics - can package and ship a service end-to-end.
-
LLM observability: designed observability for new systems with Langfuse, W&B or LangSmith - not just used existing setups.
-
Systems thinking & technical leadership: thinks in failure modes, blast radius and tail latency; writes clear design docs/ADRs; defends decisions with data; has mentored engineers with concrete impact; works effectively with non-technical stakeholders.
Nice-to-Have
-
Experience in banking, financial services or other regulated industries.
-
Hands-on fine-tuning (LoRA, QLoRA) for domain-specific tasks, especially Vietnamese or financial terminology.
-
Experience with the Qwen series in on-premise or air-gapped environments; or production work with Voicebot, OCR/VLM or other multimodal AI.
-
Open-source contributions to AI tooling (LangChain, LangGraph, vLLM, Qdrant or similar).
Nộp đơn ứng tuyển công việc này
Họ & tên bạn *
Địa chỉ email *
Số điện thoại *
Ngày tháng năm sinh *
Trình độ học vấn (Education) *
– Trình độ học vấn (Education) --Chưa tốt nghiệp (No degree)Phổ thông (High school)Trung cấpCao đẳng (College)Đại học (University/Academy)Thạc sỹ (Master)Tiến sỹ (PhD)Sau đại học (Postgraduate)Khác (Others)
Bạn biết đến cơ hội ứng tuyển này qua kênh nào? *
– Bạn biết đến cơ hội ứng tuyển này qua kênh nào? --Website tuyển dụng ACBFacebook ACBLinkedIn ACBWebsite việc làmQuảng cáo onlineNgười giới thiệuKhác
- Thông tin thêm
CV của bạn *
Click để chọn & tải lên CV của bạn
Nộp đơn ứng tuyển
Bỏ qua
→
{
“@context”: “http://schema.org”,
“@type”: “JobPosting”,
“jobLocation” : {
“@type”: “Place”,
“address”: {
“@type”: “PostalAddress”,
“addressLocality”: “No address locality”,
“addressRegion”: “No address region”,
“streetAddress”: “No street address”,
“postalCode”: “0”
}
},
“datePosted”:“2026-05-28”,
“description”:"Key Responsibilities
Architecture & Technical Design
-
Own end-to-end architecture for the team’s RAG, Agentic and Multi-Agent systems - retrieval strategy, agent orchestration patterns (Planner–Executor, Router, Verifier, Reviewer), tool-calling layer and state management.
-
Make and defend technology choices (vector DB, embedding model, orchestration framework, observability stack) with explicit trade-off analysis on capability, cost, latency and compliance.
-
Set engineering standards (coding conventions, inter-agent contracts, MCP tool schemas, minimum observability bar) and write ADRs for high-impact decisions.
-
Design reusable components (retrieval templates, MCP server skeletons, guardrail middleware, evaluation harnesses) so engineers don’t rebuild foundations for each product.
Hands-on Engineering
-
Personally write code for the hardest components: agent orchestration core, tool-calling middleware, complex retrieval, guardrail engine and tricky integrations with bank legacy systems.
-
Prototype new architectural ideas and own the hard production incidents the team can’t crack — hallucination edge cases, retrieval regressions, vLLM latency spikes, schema-breaking tool-call failures.
-
Stay hands-on: expect 40–50% of your time on direct coding. This is not a technical PM seat.
Mentorship & Code Review
-
Run high-quality code reviews that teach failure modes, cost awareness, observability and maintainability — not just check logic.
-
Mentor engineers 1-on-1 on technical growth and design decisions.
Stakeholder Representation, R&D and Reliability
-
Be the team’s technical face with Risk, Compliance, IT Infrastructure and Business - translating regulatory and business constraints into design decisions, and technical trade-offs back to stakeholders.
-
Drive technical R&D: evaluate emerging techniques, run benchmarks, separate hype from value, and contribute to the division’s technical roadmap.
-
Own production quality KPIs (hallucination rate, retrieval recall@k, tool success rate, latency, uptime); maintain evaluation frameworks that gate deployments; lead incident response and post-mortems.
-
Enforce banking-grade non-functional requirements: auditability, explainability and end-to-end traceability for internal audit and legal.
Core Requirements
Must-Have
-
Bachelor’s degree or higher in Computer Science, AI, Data Science or a related field.
-
At least 10 years of professional software engineering, with at least 3 years hands-on production work on LLM systems / RAG / Agentic AI. Shipped at least 2 AI systems to production with real users - not POCs or demos.
-
Production-grade Python: async-first (asyncio, aiohttp), Pydantic data modeling, clean modular design.
-
Real RAG / Agentic experience: debugged retrieval quality issues for real, understands cosine vs. dot product.
-
LangChain / LangGraph: fluent with the state-machine model, custom nodes and practical edge cases.
-
Vector databases & LLM serving: non-trivial production pipelines on Qdrant, Milvus or Pinecone (index schema, namespaces, metadata filtering); hands-on deployment of open-source LLMs via vLLM with concurrency, batching and quantization (AWQ/GPTQ) tuned to cost-vs-latency targets.
-
MCP & containerization: designed or implemented MCP servers / tool schemas for agent consumption in production; comfortable with Docker, Compose and Kubernetes basics - can package and ship a service end-to-end.
-
LLM observability: designed observability for new systems with Langfuse, W&B or LangSmith - not just used existing setups.
-
Systems thinking & technical leadership: thinks in failure modes, blast radius and tail latency; writes clear design docs/ADRs; defends decisions with data; has mentored engineers with concrete impact; works effectively with non-technical stakeholders.
Nice-to-Have
-
Experience in banking, financial services or other regulated industries.
-
Hands-on fine-tuning (LoRA, QLoRA) for domain-specific tasks, especially Vietnamese or financial terminology.
-
Experience with the Qwen series in on-premise or air-gapped environments; or production work with Voicebot, OCR/VLM or other multimodal AI.
-
Open-source contributions to AI tooling (LangChain, LangGraph, vLLM, Qdrant or similar).
",
“employmentType”:“Toàn thời gian”,
“hiringOrganization”: {
“@type”: “Organization”,
“name”: “Ngân hàng Á Châu - ACB”,
“sameAs” : “http://www.acbjobs.com.vn”
},
“identifier”: {
“@type”: “PropertyValue”,
“name”: “HO - AI Engineer Expert”,
“value”: “54962”
},
“title”: “HO - AI Engineer Expert”,
“validThrough”:“2026-07-31”
}
Cách thức ứng tuyển
Xem chi tiết và ứng tuyển tại: ACB
Ngày đăng: 09/06/2026
Nguồn: ACB
Đăng bởi: UB Job Crawler