Agentic Backend Engineer

The role

We are looking for an Agentic Backend Engineer to build the backend services and agent pipelines behind TaloTrace. Your main focus will be:

• Backend services in Python, built with FastAPI on Firebase and Google Cloud
• Agentic and multi-agent LLM systems that explore applications, generate test scenarios and detect bugs
• Backend systems that are automation-friendly, test-friendly and reliable enough for production use

Comfortable working with

You must be comfortable working with:

• Multi-agent systems built around LLM providers such as OpenAI, Anthropic (Claude) and Google (Gemini), mainly in Python
• Agentic coding harnesses such as Claude Code, Codex or Cursor as part of your everyday engineering workflow
• Backend workflows involving jobs, state machines, retries, data validation, evaluation loops and human review

You will own well-defined parts of the system, from design through to production, and work closely with the tech lead on the larger calls.

Examples of work you could own

You won’t do all of these. Expect to pick up one or two at a time.

Scenario generation

Turning requirements, designs and release notes into test scenarios the agents can run

Agent navigation

Improving how agents perceive and drive web, Android and iOS interfaces

Bug detection and severity

Deciding whether a finding is a real bug and how serious it is, so customers see the issues that matter first

Evidence capture

Recording the video, logs and reproduction steps that come with every finding

Issue routing

Filing findings into Jira, Linear and GitHub with enough context to act on

Run orchestration

The jobs, queues and state machines that run a test suite reliably, and recover when a step fails

Cost and caching

Run-level caching, replay and routing work between model tiers, so a full suite is fast and cheap to run

Evaluation harnesses

Measuring whether a change to an agent or a prompt made it better or worse, before it ships

You will also work on the everyday engineering problems every product has: bug fixes, backend feature upgrades, integrations, test improvements and operational reliability.

Soft skills

Above technical skills, these soft skills and values are a must.

Independence, initiative and learning

AI-assisted engineering changes fast. We value people who learn new tools quickly, solve problems logically on their own, and get real results out of ChatGPT, Claude Code, Codex, Cursor and similar agentic coding harnesses.

Communication

Raise problems early. If something is wrong with the code, the architecture, the data or the direction of a project, we want to hear it promptly and clearly.

Creativity

New ideas that simplify our processes, test pipelines or engineering automation are welcome.

Agility

Requirements change. You can follow a changed plan with good spirit.

Result oriented

Everyone is assessed on the quality and timeliness of their output.

Technical skills

We prioritise people who learn fast and have some experience over people with more experience who learn slowly. These are the skills we are looking for:

• Hands-on experience building agentic or multi-agent LLM pipelines in Python
• Proficiency with agentic coding harnesses such as Claude Code, Codex or Cursor for real implementation work, with output that holds up to review
• Experience with more than one LLM provider, such as OpenAI, Anthropic or Google
• Familiarity with tool calling, structured output, agent orchestration, memory and evaluation loops
• Experience building backend applications with FastAPI, with Flask or Django also acceptable
• Experience with Firebase and Google Cloud services such as Cloud Run, Firestore, Pub/Sub and Cloud Tasks, or similar cloud services
• A testing mindset, with care for test quality, documentation, observability and debuggability
• You treat LLM non-determinism as an engineering problem: validation, retries, fallbacks and quality checks rather than hoping for a good run

Preferred experience

Experience with the following is preferred:

• Browser or mobile automation, such as Playwright, Selenium, Appium or Maestro
• GUI agents, computer-use models or browser-driving agents
• Using LLMs to simulate user behaviour, generate tests, detect bugs or reproduce failures
• A QA or testing infrastructure background
• Setting up and maintaining CI/CD, observability and alerting
• Vector databases such as Qdrant, Weaviate or Milvus, or graph databases such as Neo4j

Years of experience

We are open to candidates with 2+ years of experience building backend systems, agentic systems or end-to-end products.

We are building a team of engineers with different levels of experience. If you have the technical skills and soft skills above, we encourage you to apply. English is not mandatory for non-leader positions.

Why join us

Real ownership early

The team is small enough that there is no process between your work and the customer. What you build ships, and you see what it does for the people using it.

Production AI agents

You work on agents and multi-agent systems that run against real applications every day, not demos.

Learn fast

You work alongside engineers who use agentic coding tools every day, and you get direct feedback on your work.

Grow with the company

As the product and team grow, so does the scope of what you can own.

Global impact

You build systems for a global market.

Benefits

• Competitive compensation package, including 100% salary during probation
• 13th-month salary and performance bonuses
• Premium health care
• 14 annual leave days and 6 sick days
• Flexible working hours
• Laptop provided (Lenovo or MacBook), or a device maintenance allowance
• Full compliance with the Vietnam Labour Code
• The opportunity to work with global stakeholders
• A dynamic, international and creative work environment
• Recognition based on skills and performance
• Strong learning and career development opportunities