高级软件工程师 - 会话人工智能
Senior Software Engineer - Conversational AI
Glia 是排名第一的银行人工智能平台，赋能社区和区域金融机构提高效率、加速贷款增长、推动存款，并提供能够战胜大银行和金融科技公司的体验。
Glia 的银行人工智能操作系统是建立在现有技术栈之上的核心智能层，激活了一个由专业代理组成的 AI 工作队伍，这些代理从银行数据、交互历史和系统记录中获取信息。这些经过银行训练的代理在语音和数字渠道（从前台到后台）自动化工作流程，从而降低运营成本并实现通用银行员模式。
Glia 凭借其坚不可摧的安全性和可靠性，赢得了 700 多家银行和信用合作社的信任。Glia 提供行业首个合同保证的无幻觉承诺。这就是为什么 Glia 客户可以迅速且自信地将银行 AI 投入使用，并从第一天起就获得可衡量的结果。有关 Glia 的更多信息，请访问 [glia.com](http://glia.com)。
### 职位名称：高级软件工程师
### 工作地点：波兰
### **了解 Vector 团队：**
Vector 团队构建直接与银行和信用合作社客户对话的 AI。当有人致电他们的信用合作社查询余额、争议费用或询问贷款时，是我们团队的代理在电话另一端。
**我们的愿景：** 构建世界上最具直觉性和可信度的面向客户的银行 AI 代理
**我们的使命：** 构建并扩展一个完全自主的语音 AI 框架，以驱动高价值的客户互动。我们统一运行时架构，减少配置摩擦，并保持现有系统完全稳定，确保客户永远不会感受到中断。
### **这项工作为何有趣：**
**具有极低延迟的语音 AI。** 人们会注意到对话中的每一个停顿。我们力求提供最优质的语音 AI 体验，这意味着要在每一步都减少毫秒级的延迟：语音识别、LLM 推理、工具调用和语音合成。延迟对我们来说是一个首要的工程约束，而不是事后考虑。
**在实际限制下的 LLM 架构。** 银行业是将 LLM 展示给客户的最严格监管领域之一。每个回复都必须安全、准确、合规且符合品牌风格，同时仍然要显得自然。这就需要为模型建立详细的控制机制：护栏、依据性、政策执行、验证层和受控的自主流程。在不牺牲速度或对话质量的前提下构建这些控制措施。
查看英文原文
**About Glia**
Glia is the #1 Banking AI platform, empowering community and regional financial institutions to create efficiencies, accelerate loan growth, drive deposits, and deliver experiences that win against megabanks and fintechs.
Glia's Banking AI Operating System is a central intelligence layer on top of existing tech stacks, activating an AI workforce of specialized agents that draw from banking data, interaction history, and integrated systems of record. These banking-trained agents automate workflows across voice and digital–from front office to back office–resulting in decreased operational costs and the Universal Banker model.
Trusted by 700+ banks and credit unions for its ironclad security and reliability, Glia delivers the industry’s first contractual no-hallucination guarantee. It’s why Glia customers quickly and confidently put Banking AI to work with measurable results from day one. More information about Glia can be found at [glia.com](http://glia.com).
### Job Title: Senior Software Engineer
### Location: Poland
### **Meet Team Vector:**
Team Vector builds the AI that talks directly to the customers of banks and credit unions. When someone calls their credit union to check a balance, dispute a charge, or ask about a loan, it’s our agent on the other end of the line.
**Our vision:** To build the world’s most intuitive and trusted customer-facing banking AI agents
**Our mission:** To build and scale a fully agentic voice AI framework that drives high-value customer interactions. We unify our runtime architecture, remove configuration friction, and keep our existing systems fully stable so customers never see a disruption.
### **Why This Work Is Interesting:**
**Voice AI with very low latency.** People notice every pause in a conversation. We aim for the best voice AI experience available, which means working to shave milliseconds off every hop: speech recognition, LLM inference, tool calls, and speech synthesis. Latency is a first-class engineering constraint for us, not an afterthought.
**LLM architecture under real constraints.** Banking is one of the most regulated places to put an LLM in front of a customer. Every response has to be safe, accurate, compliant, and on-brand, and it still has to feel natural. That calls for a detailed harness around the model: guardrails, grounding, policy enforcement, verification layers, and controlled agentic flows. Building those controls without giving up speed or conversational quality is one of the hardest and most interesting problems in applied AI right now.
**Real impact at scale.** The agents you build will handle real conversations for real people across hundreds of financial institutions. What you ship matters.
**A global team.** Team Vector works across Europe and Vancouver, Canada, and collaborates closely with teams on both continents. You’ll work with experienced engineers who care about craft and move quickly.
### **What You’ll Do:**
- Design, build, and ship the core of our agentic voice AI framework, from conversation orchestration to tool calling and runtime architecture.
- Build and evolve the harness controls (guardrails, grounding, policy checks, evaluation) that make our LLM agents trustworthy enough for banking.
- Work on latency end to end: profile, measure, and optimize every stage of the real-time voice pipeline.
- Help unify our runtime architecture and remove configuration friction so new banking use cases can launch quickly.
- Keep our existing production systems stable and reliable while we build what comes next.
- Work with product, design, and engineering teams across Europe and Vancouver to turn ambiguous problems into shipped outcomes.
- Use AI-assisted development tools daily and help set the standards for how we review and verify agent-generated code.
### **Our Tech Stack:**
- **Coding languages:** Python, Elixir, TypeScript, React (JavaScript)
- **Persistence:** Amazon RDS for PostgreSQL, DynamoDB, S3
- **Infrastructure:** AWS
- **Monitoring:** DataDog
- **CI/CD:** Jenkins, Argo CD
- **AI:** Gemini, Claude Code
- **Infrastructure as Code:** Terraform
- **Other:** Kafka, GitHub, Kubernetes, Docker
### **What We’re Looking For:**
**AI Systems Experience:** A core requirement. You have built production AI systems before and you understand LLM architecture well: prompting and context management, tool use and agentic flows, retrieval and grounding, evaluation, and the tradeoffs between latency, cost, and quality. Real-time or voice AI experience is a strong plus, and so is experience with agent SDKs (such as the Claude Agent SDK or OpenAI Agents SDK).
**Architecture & Technical Judgment:** You own the design of complex components, make sound tradeoffs, and spot risk early. When AI writes more of the code, system design becomes the scarce and highly valued skill.
**Excellent: Trusted, Fast Delivery:** You ship fast without cutting corners. You have a track record of reliable outcomes, de-risking projects, and owning quality, including verifying and standing behind agent-generated code.
**Applied AI Fluency:** You actively use tools like Claude to design, build, review, and ship. You can direct agents, verify their output, and you know where they help and where they don’t.
**Proactive:** You see problems before they’re assigned to you and act on them. You partner with PMs to take projects from idea to delivery, and you raise issues, propose solutions, and drive them forward without waiting to be asked.
**Adaptable:** You do well in a fast-paced environment where priorities, tools, and the AI landscape itself change quickly. You adjust without losing momentum.
**Curious: Learning Velocity:** You’re hungry to learn. You run at unfamiliar problems, pick up new tools and practices quickly, and share what you learn with the people around you.
**Cross-Continental Collaboration:** You communicate clearly in writing and in person, and you do well on a distributed team. You can reliably overlap at least 2 hours a day with Pacific Time (9:00–11:00 AM PT) to work closely with teammates in Vancouver.
**Leadership Instincts**: You set direction, unblock people, and raise the bar without micromanaging. Ideally, you bring prior tech lead, team lead, or mentorship experience
Glia is an equal-opportunity employer. Glia does not discriminate against any employee or applicant because of race, creed, color, religion, gender, sexual orientation, gender identity/expression, national origin, disability, age, genetic information, veteran status, marital status, pregnancy or related condition (including breastfeeding), or any other basis protected by law.
_The Glia Talent Acquisition team uses **@**_ [_**glia.com**_](http://glia.com) _and_ [_**@**_](mailto:darina.danchenko@gliatalent.com) [_**gliatalent.com**_](http://gliatalent.com) _email addresses for coordinating interviews, providing updates, and sending documents._
_Our hiring process involves an introduction, practical and team interviews, and a decision and offer. For more information, visit our_ [_Recruitment Privacy Notice page_](https://www.glia.com/eu-recruitment-privacy-notice) _or contact our talent team via **talent@glia.com**_