New opportunity

AI Engineer

Viavi Solutions Singapore Pte. Ltd. · Singapore-Techpt, Singapore

About this role

VIAVI’s CTO office seeks an AI Engineer to rapidly design, develop and deploy AI-driven solutions for telecommunications, particularly scalable 6G digital twins. The work applies generative AI, large language models and agentic architectures to challenges such as self-healing networks, automation, intelligent decision-making, customer experience and network optimization.

Responsibilities

Develop and rapidly prototype AI-based digital twins and agentic solutions for telecom workflows. Build and maintain prototypes and MVPs on cloud platforms and on-premises infrastructure.
Work with CTO leadership, AI architects, data scientists, and RF and telecom engineers to translate requirements involving 6G, 5G, Open RAN, OSS/BSS and optical networks into AI solutions.
Build pipelines for structured and unstructured telecom data; integrate hardware instruments and monitoring systems with cloud-based AI.
Apply model evaluation, validation, explainability and monitoring practices. Keep current with AI/ML research and telecom AI applications. The role may involve applied research, scientific publications, patent applications and coordination across teams.

Required qualifications

A record of shipping AI solutions from concept through deployment and owning the ML development lifecycle; experience developing generative AI applications, including agent-based architectures.
Strong Python skills; hands-on experience with TensorFlow or PyTorch, dataset curation, LLM fine-tuning, and generative AI tools or agentic frameworks such as LlamaIndex, CrewAI, LangChain, Hugging Face and OpenAI APIs.
Hands-on experience across the MLOps workflow, including data processing, training infrastructure and deployment infrastructure. A master’s degree in Computer Science, Electrical Engineering, Data Science or a related field is preferred.

Preferred qualifications

Familiarity with RF network architectures, protocols and telecom data; network optimization, anomaly detection or predictive analysis; RF planning, network slicing, Open RAN or OSS/BSS; and data engineering technologies such as Kafka or Spark Streaming. Open-source AI contributions or participation in AI communities are also preferred.

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

Skills for this role

Generative AILarge Language ModelsAgentic AIDigital twinsPythonTensorFlowPyTorchLLM fine-tuningDataset curationLlamaIndexCrewAILangChainHugging FaceOpenAI APIsMLOpsData pipelinesModel evaluationModel validationExplainabilityModel monitoringCloud platformsTelecommunications6G5GOpen RANOSS/BSSOptical networksRF network architecturesNetwork optimizationAnomaly detection service. But no exact closing evidence. Wait accidental?

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

Prototype telecom digital twins and agentic AI
Rapidly develop AI-based digital twins and agentic solutions for telecom workflows, with a focus on scalable 6G applications. Maintain prototypes and MVPs across cloud and on-premises infrastructure.
Own AI solutions through deployment
The role calls for shipping AI solutions from concept to deployment and owning the ML lifecycle, including data processing, training infrastructure and deployment infrastructure.
Translate telecom needs into AI designs
Work with CTO leadership, AI architects, data scientists, and RF and telecom engineers to turn needs involving 6G, 5G, Open RAN, OSS/BSS and optical networks into AI solutions.
Connect telecom data and instruments
Design pipelines for structured and unstructured telecom data, and integrate hardware instruments and monitoring systems with cloud-based AI.
Evaluate and monitor model behavior
Apply model evaluation, validation, explainability and monitoring practices. The role also calls for staying current with AI/ML and telecom AI research; publications and patent applications are possible, not stated requirements.

Likely challenges to think through

Speed versus credible validation
Likely preparation hypothesis: fast prototyping may compete with the need to validate and explain digital-twin or agentic outputs. Prepare an example of how you would set an MVP scope, choose evaluation measures and decide what needs monitoring before deployment.
Integration across unlike systems
Likely preparation hypothesis: combining structured and unstructured telecom data, hardware instruments, monitoring systems, cloud AI and on-premises infrastructure may make interfaces and data quality difficult. Prepare a pipeline design that shows how you would check inputs, handle failures and trace outputs.
Turning network goals into a tractable AI task
Likely preparation hypothesis: broad goals such as self-healing networks or optimization may need narrowing with telecom specialists before a useful prototype can be built. Prepare to explain how you would define one workflow, its available data and a measurable outcome with those collaborators.

Practice your answers

Technical practice

How would you design an AI-based digital twin for a telecom network?

What to coverThe role focuses on scalable 6G digital twins and rapid prototyping. Preparation suggestion: explain what network state the twin would represent, how you would keep it aligned with observed data, and how you would test its usefulness before scaling it.

When would you use an agentic architecture in a telecom workflow?

What to coverThe posting calls for generative AI applications and agent-based solutions within telecom workflows. Preparation suggestion: distinguish tasks suited to an agent from simpler automation, and discuss how you would check its decisions before they affect a network.

How would you build a pipeline for structured and unstructured telecom data?

What to coverDesigning these pipelines is a stated responsibility. Preparation suggestion: cover ingestion, dataset quality, transformations and how you would make data usable for training or inference; Kafka or Spark Streaming are listed as preferred knowledge, not required tools.

How would you connect hardware instruments and monitoring systems to cloud-based AI?

What to coverThe role explicitly includes developing heterogeneous systems that make this connection. Preparation suggestion: describe interface and data-format choices, timing and reliability concerns, and how you would troubleshoot inconsistent readings across systems.

What would you consider when moving an AI prototype between cloud and on-premises infrastructure?

What to coverThe posting calls for building and maintaining prototypes and MVPs in both environments. Preparation suggestion: compare deployment, data movement, performance and operational constraints without assuming a particular platform or architecture.

How would you curate a dataset and fine-tune an LLM for a telecom use case?

What to coverDataset curation and LLM fine-tuning are listed as hands-on requirements. Preparation suggestion: discuss data selection and quality, a baseline to compare against, and how you would detect whether fine-tuning improves the intended task.

How would you evaluate, explain and monitor an AI solution after deployment?

What to coverModel evaluation, validation, explainability and monitoring are explicit responsibilities. Preparation suggestion: tie measures to the intended telecom task, distinguish pre-deployment checks from live monitoring, and explain how you would investigate a decline in performance.

How would you take a Python ML prototype through the full development lifecycle?

What to coverThe role requires Python, TensorFlow or PyTorch, and hands-on ownership from data processing through training and deployment infrastructure. Preparation suggestion: walk through reproducibility, testing, deployment and maintenance decisions, including what you would change between a prototype and a shipped solution.

How would you translate a telecom expert’s requirements into an AI problem?

What to coverThe role involves working with specialists across 6G, 5G, Open RAN, OSS/BSS and optical networks. Preparation suggestion: explain how you would clarify the network workflow, available data, desired outcome and technical constraints before choosing a model or approach.

How would you assess an AI approach to network anomalies or self-healing?

What to coverThe posting names self-healing networks and network optimization as target challenges, with anomaly detection listed among preferred experience. Preparation suggestion: compare detection with deciding what action to take, and discuss false alarms, validation and the risks of acting on uncertain predictions.

Behavioural practice

Tell me about a time you owned an AI solution from concept to deployment.

What to coverShipping AI solutions across the ML lifecycle is a stated requirement. Preparation suggestion: use a real situation, specify your decisions and actions at each stage, and give the observed result without overstating your individual contribution.

Describe a time you worked with domain experts to resolve unclear requirements.

What to coverThe role involves collaboration with CTO leadership, AI specialists, data scientists, and RF and telecom engineers. Preparation suggestion: describe the situation, how you elicited and reconciled requirements, and what changed as a result.

Tell me about a time you had to prototype quickly while managing technical trade-offs.

What to coverRapid prototyping and maintaining MVPs are central to the advertised work. Preparation suggestion: explain the time constraint, what you chose to simplify or validate first, and the outcome and limitations of the prototype.

Describe a time you investigated whether a new technical idea was feasible.

What to coverThe role summary includes feasibility investigations and applied research. Preparation suggestion: give a real example of the question you tested, the evidence you gathered, your actions and the resulting decision.

Tell me about a time evaluation findings led you to revise an AI solution.

What to coverThe posting emphasizes model validation, explainability and monitoring. Preparation suggestion: set out the situation, the issue the evidence revealed, how you responded and the measurable or observable result.

Sharpen your knowledge

Topics to prepare

Agentic AI and 6G digital twins
Sketch how an agent could use observations from a digital twin to recommend a network action, then mark where a person or system would validate it. The posting explicitly calls for agentic telecom workflows and scalable 6G digital-twin solutions; this sketch is a preparation exercise, not a specified design.
Telecom systems and data
Draw a high-level map of 5G, 6G, Open RAN, OSS/BSS and optical networks, noting where RF data and monitoring signals might enter an AI workflow. The role involves translating needs with telecom specialists; RF concepts and telecom data types are listed as preferred knowledge.
Evaluation, validation and explainability
Draft an evaluation plan for a proposed network-anomaly assistant: define what to test, how to check incorrect recommendations, and what explanations an engineer would need. The posting explicitly names model evaluation, validation, explainability and monitoring; the anomaly scenario is a suggested exercise.
ML lifecycle and deployment trade-offs
Outline the stages from dataset preparation and training through prototype, deployment and monitoring, comparing what would need attention on cloud versus on-premises infrastructure. End-to-end ML ownership and MLOps are required, and prototypes on both infrastructure types are part of the role.
Network optimization use cases
Compare anomaly detection, predictive analysis and self-healing as possible network workflows, identifying the data, decision and feedback each would need. Network optimization and self-healing are stated applications; anomaly detection and predictive analysis appear under preferred skills.

Skills to revise

Python model prototyping
Implement a small Python training-and-inference prototype in either PyTorch or TensorFlow, and record how you would test its outputs. Strong Python skills and hands-on experience with one of these frameworks are required; the prototype is a suggested practice task.
Dataset curation and LLM fine-tuning
Prepare a small, clearly labeled example dataset, document exclusions and train/validation separation, then rehearse a fine-tuning run and compare outputs against a baseline. Dataset curation and LLM fine-tuning are explicit requirements; the example dataset is for practice.
Agentic application construction
Build a limited agent workflow with one framework named in the posting, such as LangChain or CrewAI, and test how it handles an unavailable or contradictory observation. The role requires GenAI and agent-based application experience; no particular framework is mandated.
Telecom data pipelines and integration
Prototype a pipeline that combines sample structured network measurements with unstructured notes, then specify how instrument or monitoring-system data would reach cloud-based AI. Pipelines for both data forms and hardware/monitoring integration are stated responsibilities; sample inputs are a suggested substitute for live systems.
MLOps deployment and monitoring
Package a prototype for a repeatable deployment exercise, record its data-processing and training steps, and define checks for output quality after deployment. The posting requires hands-on MLOps across processing, training and deployment infrastructure, alongside model monitoring.

Your preparation checklist

Prepare a shipped-AI project walkthrough
The posting asks for concept-to-deployment ownership. Prepare one project timeline showing your decisions, Python and TensorFlow or PyTorch work, deployment approach, results, and what you maintained afterward.
Sketch a telecom digital-twin design
As a practice exercise, diagram a prototype for network anomaly detection or optimization. Show what the twin represents, where an agentic workflow fits, and how you would test an MVP on cloud or on-premises infrastructure.
Document a GenAI implementation
The role calls for agent-based applications, dataset curation, and LLM fine-tuning. Prepare a concrete example with data choices, model or framework choices, evaluation results, and limitations; identify any work you have not done.
Map the data and instrument interfaces
Draft a sample flow for structured and unstructured telecom data, from monitoring systems or hardware instruments through processing to cloud-based AI. Mark assumptions and questions for RF and telecom engineers.
Prepare an evaluation and MLOps plan
The posting names validation, explainability, monitoring, and the full MLOps workflow. Outline test data, success and failure measures, training and deployment infrastructure, and checks after release.
Review telecom concepts relevant to the design
Prepare a brief glossary of 5G/6G, RF, Open RAN, OSS/BSS, optical networks, and network slicing. These appear in the role or preferred skills; note which concepts would need clarification from subject-matter experts.
Write questions for the CTO office
Ask how the team selects digital-twin use cases, obtains telecom and instrument data, measures prototype success, and divides work among AI architects, data scientists, and RF engineers. The posting describes these collaborators and activities but not their specific processes.

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…