Senior Software Engineer - Conversational AI (Python)
About this role
This role addresses the challenge of building highly reliable, low-latency conversational AI agents capable of handling complex, regulated customer interactions within the financial sector. Glia is focused on creating a central intelligence layer that automates workflows for banking agents, aiming to reduce operational costs while maintaining high standards of security and accuracy. The core problem this hire will solve is engineering the core of the agentic voice AI framework that connects directly with customers of banks and credit unions.
The Senior Software Engineer will be responsible for designing, building, and shipping this framework, focusing heavily on the real-time performance constraints of voice interactions. This involves deep work on the LLM architecture under operational constraints, ensuring agents are safe, compliant, and maintain high conversational quality. Furthermore, the engineer will be instrumental in optimizing the entire voice pipeline to minimize latency across speech recognition, inference, and synthesis.
This position is ideal for an engineer who thrives on the intersection of applied AI, robust system design, and high-stakes environments. A successful candidate will enjoy tackling the difficult problems of building safety harnesses for LLMs while simultaneously optimizing for speed and delivering tangible impact to real users across hundreds of financial institutions.
What you will do
- Design, build, and deploy the core components of the agentic voice AI framework, encompassing conversation orchestration and tool calling mechanisms.
- Develop and iterate on the control mechanisms, such as guardrails, grounding strategies, and policy checks, to ensure LLM agents are suitable for banking applications.
- Optimize the end-to-end latency of the real-time voice pipeline by profiling and tuning every stage, including speech recognition and LLM inference.
- Contribute to unifying the runtime architecture to reduce configuration friction, enabling the rapid launch of new banking use cases.
- Maintain the stability and reliability of existing production systems while advancing new AI capabilities.
- Collaborate with product and design teams across different geographies to translate ambiguous requirements into shipped solutions.
- Utilize AI-assisted development tools daily to help establish standards for reviewing and verifying agent-generated code.
What you need
- A strong background in AI systems is required, demonstrated by prior experience building production AI systems and a solid understanding of LLM architecture, including prompting, context management, tool use, and agentic flows.
- Proficiency in understanding the trade-offs between latency, cost, and quality when designing AI solutions is essential.
- Experience with real-time or voice AI systems is considered a strong asset, as is familiarity with agent SDKs like the Claude Agent SDK or OpenAI Agents SDK.
- Demonstrated ability in architecture and technical judgment is necessary to own complex component design and make sound trade-offs, especially when AI contributes significantly to code generation.
- A track record of delivering trusted, fast results and successfully de-risking projects by owning quality and verification processes is required.
Nice to have
- Real-time or voice AI experience
- Experience with agent SDKs (such as the Claude Agent SDK or OpenAI Agents SDK)
Benefits
- A global team working across Europe and Vancouver, Canada
Skills and tech stack
Questions candidates ask
What is the remote policy for this role?
The location for this role is listed as Estonia - Remote. The team works across Europe and Vancouver, Canada, indicating a distributed setup.
What is the compensation structure?
The posting does not specify the salary for this position.
What is the primary technology stack used?
The tech stack includes coding languages like Python, Elixir, TypeScript, React (JavaScript), and various cloud services such as AWS, along with databases like Amazon RDS for PostgreSQL and DynamoDB.
What kind of team structure is involved?
The work is done by Team Vector, which collaborates closely with teams on both Europe and Vancouver continents.