New opportunity

Agentic AI Engineer

Bitdeer Technologies Group · Singapore

About this role

Bitdeer Technologies Group seeks an Agentic AI Engineer for its AI Cloud business, with locations listed as Singapore, SG and Penang, MY. The role develops autonomous systems to improve NeoCloud engineering and business workflows, including infrastructure resilience and operations.

Responsibilities

Design scalable multi-agent architectures using planners, executors and judges. Connect agents to internal services, SaaS platforms and APIs so they can initiate actions such as node cordoning, checkpointing and hardware remediation.
Build agents that generate and refactor code, analyze telemetry and correct system issues in real time. Design short- and long-term agent memory for long-running tasks using protocols such as Model Context Protocol (MCP).
Develop Retrieval-Augmented Generation (RAG) pipelines grounded in telemetry and observability data. Apply LLM-as-a-Judge evaluation and guardrails for performance, reliability and security compliance.
Establish Harness engineering practices for deployment, governance and lifecycle management of agentic systems, using CI/CD and feature flagging. Work with Kubernetes and infrastructure engineering teams to integrate agents into the platform control plane.

Qualifications

Bachelor’s or Master’s degree in Computer Science, Artificial Intelligence, Software Engineering or a related field; a minimum of 6+ years of software engineering experience; and deep expertise in production-grade Python and TypeScript.
Extensive experience with agentic frameworks such as LangGraph or CrewAI and frontier LLMs such as OpenAI models or Anthropic Claude. Demonstrated ability to turn ambiguous business processes into reusable AI systems.
Experience deploying AI features in managed runtimes and cloud-native environments, including AWS, Azure, GCP or private clouds. Familiarity with Kubernetes internals, infrastructure automation and CI/CD tools such as Jenkins, GitHub Actions and Harness.
Strong understanding of distributed systems, telemetry, observability tools such as Prometheus and OpenTelemetry, and hardware-level metrics. Leadership, communication and cross-functional collaboration skills are also sought.

The posting describes training, mentoring, developmental opportunities and welfare benefits; it does not specify compensation.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

Multi-agent systemsAI agentsPythonTypeScriptLangGraphCrewAILarge Language ModelsOpenAIClaudeModel Context ProtocolRetrieval-Augmented GenerationLLM-as-a-JudgeAgent evaluationAI guardrailsKubernetesInfrastructure automationCI/CDFeature flaggingHarnessJenkinsGitHub ActionsAWSAzureGCPDistributed systemsPrometheusOpenTelemetryTelemetry analysisObservabilityAPI integration

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Scalable agent architecture
The role designs planner, executor and judge agents for end-to-end NeoCloud workflows, turning ambiguous processes into reusable AI systems. It also calls for short- and long-term memory to maintain continuity across long-running tasks, using protocols such as MCP.
Infrastructure action integration
Agents must connect to internal services, SaaS platforms and APIs to initiate actions such as node cordoning, checkpointing and hardware remediation. The role also works with Kubernetes and infrastructure teams to integrate agents into the platform control plane.
Telemetry-grounded self-correction
The posting calls for agents that analyze telemetry and correct system issues in real time, supported by RAG pipelines grounded in observability data. Distributed-systems knowledge and familiarity with hardware-level metrics matter here.
Evaluation and guardrails
The role is responsible for LLM-as-a-Judge evaluation and guardrails aimed at agent performance, reliability and security compliance.
Agent deployment and governance
The role establishes company-wide Harness engineering practices for deploying, governing and managing agentic systems through their lifecycle, using CI/CD and feature flagging. Cross-functional leadership and production-grade Python and TypeScript are also emphasized.

Likely challenges to think through

Autonomy versus operational risk
Likely preparation challenge, not a confirmed Bitdeer problem: letting an agent act quickly on infrastructure could conflict with limiting the impact of a wrong action. Prepare to explain how you would define action boundaries, validate decisions and handle failures for actions such as node cordoning.
Continuity versus changing system state
Likely preparation challenge: a long-running agent needs memory, but stored context or retrieved telemetry may no longer reflect current infrastructure state. Prepare to discuss how an agent would check data freshness, preserve task history and re-evaluate before acting.
Rapid iteration versus dependable rollout
Likely preparation challenge: improving agent behavior quickly can conflict with reliable, governed deployment. Prepare to describe how evaluation results, CI/CD and feature flags could support staged changes and reveal regressions without assuming a specific Bitdeer process.

Practice your answers

Technical practice

How would you design a scalable planner–executor–judge architecture for a NeoCloud workflow?

What to coverThe posting calls for multi-agent architectures that automate engineering and business workflows. Preparation suggestion: Explain agent responsibilities, how state passes between them, and how you would limit loops and recover from failed steps.

How would you let an agent trigger node cordoning or hardware remediation safely?

What to coverThe role includes connecting agents to internal services, SaaS platforms and APIs for infrastructure actions. Preparation suggestion: Discuss authorization, action validation, idempotency and when an action should be blocked or escalated.

How would you use telemetry to detect and correct a system issue in real time?

What to coverThe posting describes agents that analyze telemetry and self-correct issues, and seeks knowledge of Prometheus, OpenTelemetry and hardware-level metrics. Preparation suggestion: Walk through signal selection, diagnosis, confidence in a proposed correction and verification that the system recovered.

How would you build an agent that generates or refactors production code?

What to coverCode generation and refactoring are stated duties, alongside production-grade Python and TypeScript expertise. Preparation suggestion: Explain how you would constrain changes, test behavior and review a proposed edit before it affects a running system.

How would you preserve agent context across a long-running infrastructure task?

What to coverThe role calls for short- and long-term agent memory and mentions Model Context Protocol (MCP). Preparation suggestion: Distinguish working state from durable history, and address stale context, resumability and access to sensitive information.

How would you ground an agent’s decision in telemetry using RAG?

What to coverThe posting explicitly calls for RAG pipelines using reliable data from telemetry and observability stacks. Preparation suggestion: Describe retrieval and freshness choices, how the agent cites or checks retrieved evidence, and what it should do when evidence is missing or contradictory.

How would you evaluate an agent using LLM-as-a-Judge without relying on it alone?

What to coverThe role calls for LLM-as-a-Judge evaluation and guardrails for performance, reliability and security compliance. Preparation suggestion: Discuss representative test cases, judge limitations, measurable outcomes and independent checks for unsafe actions.

How would you deploy and govern changes to an agentic system?

What to coverThe posting assigns company-wide Harness engineering practices for deployment, governance and lifecycle management, using CI/CD and feature flagging. Preparation suggestion: Explain staged rollout, regression checks, monitoring and rollback, including trade-offs between release speed and operational risk.

What changes when agents integrate with a Kubernetes platform control plane?

What to coverThe role involves collaboration with Kubernetes and Infrastructure Engineering teams to improve platform resilience. Preparation suggestion: Discuss reconciliation, distributed-system failure modes and boundaries on automated control-plane actions.

How would you run an agentic service reliably in a managed or cloud-native environment?

What to coverThe posting seeks experience deploying AI features in managed runtimes and cloud-native environments, including public or private clouds. Preparation suggestion: Cover service availability, scaling, observability and dependency failures without assuming a particular cloud provider is required.

Behavioural practice

Tell me about a time you turned an ambiguous process into a reusable system.

What to coverThe posting seeks experience translating ambiguous business processes into robust, reusable AI logic. Preparation suggestion: Use a real situation, the steps you took to clarify the process and design a solution, and the resulting outcome.

Describe a time you coordinated with infrastructure specialists on a risky integration.

What to coverThe role explicitly involves collaboration with Kubernetes and Infrastructure Engineering teams. Preparation suggestion: Give a real situation, your actions to resolve differing priorities or ownership boundaries, and the result.

Tell me about a time you helped a team adopt an unfamiliar engineering approach.

What to coverThe posting seeks leadership and communication that bridge traditional software engineering and autonomous AI. Preparation suggestion: Describe a real situation, how you explained trade-offs and addressed concerns, and what changed as a result.

Describe a time you improved reliability after a production issue.

What to coverThe role emphasizes reliable software, observability and infrastructure resilience. Preparation suggestion: Use a real situation, your diagnostic and corrective actions, and the measured or observed result; distinguish your contribution from the team’s.

Tell me about a time you balanced delivery speed against operational risk.

What to coverThe posting describes a high-velocity engineering environment and calls for deployment governance and feature flagging. Preparation suggestion: Explain a real situation, the trade-off you made and safeguards you put in place, and the outcome.

Sharpen your knowledge

Topics to prepare

Multi-agent architecture
The posting calls for scalable planners, executors and judges across NeoCloud workflows. As a revision exercise, diagram how those roles would coordinate on a hypothetical infrastructure incident, including when an action is rejected or retried.
Agent memory and MCP
The role specifies short- and long-term memory for long-running tasks using protocols such as MCP. As a revision exercise, trace what context an agent needs before and after a task interruption, and distinguish transient state from retained history.
Telemetry-grounded RAG
The posting requires RAG grounded in telemetry and observability data. As a revision exercise, sketch how an agent would retrieve relevant incident evidence, assess its freshness and cite it before recommending an action.
Evaluation and security guardrails
The role names LLM-as-a-Judge evaluation and guardrails for performance, reliability and security compliance. As a revision exercise, define success and failure criteria for a hypothetical remediation decision, including cases where the agent must not act.
Distributed infrastructure and Kubernetes
The posting seeks distributed-systems understanding and Kubernetes internals for control-plane integration. As a revision exercise, explain the possible system effects of node cordoning during a partial failure and what an agent would need to verify first.
Agent lifecycle governance
The role calls for Harness engineering practices, CI/CD and feature flagging to manage agent deployment and lifecycle. As a revision exercise, outline a staged release and rollback plan for a hypothetical agent change; this plan is preparation, not a stated Bitdeer process.

Skills to revise

Production-grade Python and TypeScript
Both languages are explicitly sought. As a revision exercise, implement a small typed agent-to-service interface in each language, with input validation, error handling and tests.
Agent workflow implementation
The posting cites LangGraph and CrewAI as framework examples. As a revision exercise, use one to build a small planner–executor–judge workflow and test how it handles a failed execution.
Safe API action integration
Agents are expected to connect to internal services, SaaS platforms and APIs for infrastructure actions. As a revision exercise, connect an agent to a mock node-cordoning API and test authorization checks, repeated requests and failed responses; the mock is a practice suggestion.
Observability data handling
The role names Prometheus, OpenTelemetry and hardware-level metrics. As a revision exercise, inspect sample metrics and traces, then produce an evidence summary that distinguishes an observed symptom from a proposed cause.
RAG and evaluation testing
The posting requires telemetry-grounded RAG and LLM-as-a-Judge evaluation. As a revision exercise, build a small retrieval test set from sample observability records and score whether generated answers use the retrieved evidence correctly.
CI/CD and feature-flagged delivery
The role names Jenkins, GitHub Actions and Harness as CI/CD examples and calls for feature flagging. As a revision exercise, draft a pipeline for testing an agent change, enabling it behind a flag and disabling it after a failed evaluation; choose one named CI/CD tool for practice.

Your preparation checklist

Select a production project walkthrough
If available, prepare one example showing how you turned an ambiguous process into reusable AI logic. Bring an architecture diagram, your specific contributions, and evidence of production-grade Python or TypeScript, agent frameworks, and cloud deployment.
Sketch a NeoCloud multi-agent design
As a proposed design exercise, diagram planner, executor, and judge roles for a NeoCloud workflow. Explain task boundaries, coordination, failure recovery, and how you would assess scalability.
Trace an infrastructure action safely
Walk through a hypothetical agent-triggered node cordon or checkpoint: the API integration, authorization, validation, audit trail, and rollback. Identify where Kubernetes and infrastructure engineers would need to collaborate.
Design telemetry-grounded decisions
Prepare a data-flow sketch using observability and hardware metrics to support RAG-based decisions. Explain short- and long-term memory for a long-running task and where MCP might fit.
Build an evaluation plan
Draft example tests for agent reliability, performance, and security compliance, including failure cases and an LLM-as-a-Judge rubric. Distinguish automated judgments from checks needed before an agent acts.
Outline a deployment and governance approach
Propose how CI/CD, feature flags, monitoring, and lifecycle controls could support agent rollout. Relate the approach to the posting’s Harness engineering responsibility without assuming Bitdeer’s current tooling.
Prepare cross-functional leadership examples
If available, select examples of decisions made with software, Kubernetes, or infrastructure teams. Be ready to explain trade-offs, communication, and measurable outcomes without overstating your role.
Write targeted employer questions
Ask which NeoCloud workflows are priorities, what authority agents may have over infrastructure actions, how telemetry quality is assessed, and how the team defines readiness for control-plane integration.

Availability last checked: 4 Nov 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in Singapore. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…