New opportunity

Solutions Architect, Speech AI

SG02 NVIDIA SG Dev Pte. Ltd. · Singapore

About this role

NVIDIA seeks a Solutions Architect in Singapore to help customers and partners across South East Asia build, deploy and scale speech AI solutions. The role is the first line of technical expertise for customers and partners, combining solution design, technical validation, deployment support and enablement.

Responsibilities

Work with customers, global systems integrators, independent software vendors and internal account teams on voice AI applications, including real-time transcription, multilingual speech recognition, speech synthesis, speech translation and conversational AI.
Validate speech AI models and workflows involving multilingual configurations, code-switching and streaming recognition. Translate customer requirements into deployable architectures, technical feedback and product roadmap input with field, product and engineering teams.
Assist with deploying NVIDIA Riva, Nemotron and speech NIM microservices in cloud, data center and edge environments, addressing streaming, scaling and performance. Deliver demos, workshops, technical reviews and partner enablement; support troubleshooting, project coordination and customer relationships.

Required qualifications

A BS, MS or PhD in Computer Science, Electrical Engineering, Computer Engineering, Physics, Mathematics, Machine Learning or a related field, or equivalent experience; 5+ years of experience building and deploying speech, multimodal AI or LLM systems.
Strong Python and PyTorch skills; hands-on experience with ASR, TTS, speech understanding, speech translation, audio-language models or multimodal LLMs. Understand production trade-offs among latency, quality and scale, and be familiar with Docker, Kubernetes, Helm or equivalent infrastructure tools.
Ability to explain end-to-end speech AI pipelines covering model selection, customization, deployment, observability and application integration, with strong communication across customers, partners, product management and engineering.

Preferred qualifications include experience with NeMo Speech, Riva, NIM, Nemotron or Parakeet; multilingual deployments and language-specific adaptation; speech-to-speech systems; AI agent benchmarking for quality, latency, reliability, safety and user experience; and production voice AI applications with enterprise partners in contact centers, digital humans, assistants or industry-specific workflows. NVIDIA mentions competitive salaries and a benefits package but provides no salary figure.

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

Skills for this role

PythonPyTorchAutomatic Speech Recognition (ASR)Text-to-Speech (TTS)Speech UnderstandingSpeech TranslationAudio-Language ModelsMultimodal LLMsConversational AIDockerKubernetesHelmNVIDIA NeMo SpeechNVIDIA RivaNVIDIA NIMNVIDIA NemotronNVIDIA ParakeetMultilingual Speech RecognitionCode-SwitchingStreaming Speech RecognitionModel DeploymentObservabilityPerformance BenchmarkingSolution ArchitectureTechnical Communication

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

Design end-to-end voice AI solutions
The role works with customers, systems integrators, software vendors and account teams on transcription, speech synthesis, translation and conversational AI. The posting calls for architectures that cover model selection, customization, deployment, observability and application integration.
Validate multilingual and streaming workflows
A stated duty is technical validation of NVIDIA speech AI models, including multilingual configurations, code-switching and streaming recognition.
Support production deployment
The architect assists with Riva, Nemotron and speech NIM deployments across cloud, data center and edge environments, with attention to streaming, scaling and performance.
Turn customer needs into technical direction
The posting calls for working with field, product and engineering teams to convert customer requirements into deployable architectures, technical feedback and roadmap input.
Enable customers and partners
Demos, workshops, technical reviews, troubleshooting and partner enablement are part of the first-line technical role, alongside customer relationship management.

Likely challenges to think through

Balancing recognition quality with responsiveness
Preparation hypothesis, not a confirmed employer problem: multilingual and code-switched streaming speech may make quality, latency and scale difficult to optimize together. Prepare to explain how you would validate representative speech scenarios, measure the trade-offs and justify an architecture choice.
Making one solution deployable across environments
Preparation hypothesis, not a confirmed employer problem: cloud, data center and edge deployments may have different performance and operational constraints. Prepare to walk through deployment choices, scaling, observability and troubleshooting for a speech AI pipeline without assuming the same configuration fits every setting.
Reconciling stakeholder needs with technical limits
Preparation hypothesis, not a confirmed employer problem: customer goals, partner implementation needs and product capabilities may not align neatly. Prepare a concise way to turn requirements into a feasible design, explain trade-offs in a technical review and convey actionable feedback to product and engineering.

Practice your answers

Technical practice

How would you design an end-to-end real-time transcription pipeline for a customer application?

What to coverThe posting calls for model selection, customization, deployment, observability, and application integration. Suggested answer focus: trace audio from input through streaming ASR to the application, explaining where you would measure quality and latency.

How would you validate multilingual ASR when speakers switch languages mid-utterance?

What to coverThe role includes multilingual configurations, code-switching, and language-specific adaptation. Suggested answer focus: describe representative test audio, language-level error analysis, and how results would guide configuration or customization.

How would you troubleshoot delayed or inaccurate streaming recognition?

What to coverThe posting names streaming workflows, troubleshooting, and production performance. Suggested answer focus: separate audio-input, model, and serving issues, then explain how you would test a fix without overlooking recognition quality.

How would you connect speech recognition, translation, and synthesis in a voice application?

What to coverThe advertised scenarios include speech translation, speech synthesis, and conversational AI. Suggested answer focus: explain component boundaries and how you would assess the combined system's latency and output quality.

How would you use Python and PyTorch to validate or customize a speech model?

What to coverStrong Python skills, PyTorch experience, and speech-model validation are stated requirements. Suggested answer focus: describe a reproducible evaluation or customization workflow, including baseline comparisons and checks for regressions.

What would change when deploying Riva, Nemotron, or speech NIM microservices in cloud, data center, and edge environments?

What to coverThe role supports these deployments and their streaming, scaling, and performance considerations. Suggested answer focus: compare environment constraints and explain how you would validate that the chosen architecture meets customer requirements.

How would you plan infrastructure for a speech service whose concurrent usage is growing?

What to coverThe posting mentions Docker, Kubernetes, Helm or equivalent tooling, alongside production scaling. Suggested answer focus: discuss deployment configuration, capacity testing, and the trade-off between responsiveness and resource use.

How would you measure latency, quality, and scale after a speech AI service goes live?

What to coverProduction trade-offs and pipeline observability are explicit requirements. Suggested answer focus: identify useful service and model-level measurements, and explain how you would investigate a change in one without assuming the others are unaffected.

How would you benchmark a conversational voice AI agent?

What to coverBenchmarking agents for quality, latency, reliability, safety, and user experience is listed as preferred experience. Suggested answer focus: propose representative interactions and explain how you would interpret trade-offs across those dimensions.

How would you turn a customer's voice AI requirements into a technically validated architecture?

What to coverThe role translates customer requirements into deployable architectures, technical feedback, and roadmap input. Suggested answer focus: explain how you would clarify the use case, compare options, validate key assumptions, and distinguish a deployment issue from product feedback.

Behavioural practice

Tell me about a time you resolved differing technical priorities among a customer, partner, and internal team.

What to coverThe role works across customers, partners, and internal account teams. Suggested answer focus: use a real situation, your actions to clarify priorities and reach a decision, and the resulting outcome.

Describe a time you turned field observations into actionable feedback for product or engineering colleagues.

What to coverThe posting calls for technical feedback and roadmap input with field, product, and engineering teams. Suggested answer focus: explain the situation, how you substantiated and communicated the feedback, and what resulted.

Tell me about a difficult deployment or production issue you helped troubleshoot with a customer.

What to coverDeployment support, troubleshooting, and customer relationships are stated duties. Suggested answer focus: describe the issue, your diagnostic and communication actions, and the result without overstating your individual contribution.

Describe a workshop or technical review where you had to make complex material useful to others.

What to coverThe role delivers demos, workshops, technical reviews, and partner enablement. Suggested answer focus: give the audience and situation, explain how you adapted the session, and describe evidence of its outcome.

Tell me about a time you coordinated a technical project across multiple organizations.

What to coverProject coordination and collaboration with customers, integrators, software vendors, and internal teams are part of the role. Suggested answer focus: outline the situation, how you managed responsibilities or blockers, and the resulting progress.

Sharpen your knowledge

Topics to prepare

End-to-end speech AI architecture
Sketch a customer application from audio input through model selection, customization, deployment, observability and integration; explain each design choice. The posting explicitly requires building and explaining this pipeline.
Voice AI model choices
Compare when to use ASR, TTS, speech understanding, speech translation, audio-language models and multimodal LLMs in two voice-application scenarios. The role calls for hands-on experience with these approaches and covers transcription, synthesis, translation and conversational AI.
Multilingual and streaming recognition
Walk through how you would validate a streaming recognizer on multilingual speech and code-switching, including what failures you would inspect. These are explicit technical-validation responsibilities; language-specific adaptation is preferred.
Production trade-offs
For a real-time transcription scenario, explain how a change intended to improve quality might affect latency and scale, and what you would measure before deployment. The posting explicitly asks for experience with these trade-offs.
Deployment across environments
Compare a proposed speech-service architecture in cloud, data center and edge settings, identifying streaming, scaling and performance considerations. The role assists deployments of Riva, Nemotron and speech NIM microservices in all three environments.

Skills to revise

Python and PyTorch implementation
As a preparation exercise, write and explain a small Python/PyTorch speech-model inference or validation workflow, including how you would inspect its outputs. Strong Python and PyTorch experience is explicitly required.
Speech workflow validation
Prepare a repeatable test plan with audio samples covering streaming recognition, multiple languages and code-switching; specify the results and failure cases you would record. The role explicitly involves technical validation of these workflows.
Containerized deployment troubleshooting
Rehearse diagnosing a speech microservice that starts but has poor streaming performance, checking deployment configuration and service behavior. Docker, Kubernetes, Helm or equivalent tooling is required familiarity, and deployment support and troubleshooting are listed duties.
Customer-facing architecture review
Practice turning a customer's voice-application requirements into an architecture diagram and a short technical review, then state which questions need input from product or engineering. Translating requirements and delivering technical reviews are explicit duties.
Demo and workshop delivery
Outline a concise speech AI demo and a partner workshop segment that explain the workflow, deployment considerations and limitations to a technical audience. Demos, workshops and partner enablement are stated responsibilities.
Agent benchmark design
As optional preparation for a preferred qualification, draft a benchmark plan for a conversational voice agent covering quality, latency, reliability, safety and user experience. The posting lists experience benchmarking AI agents on these dimensions as a way to stand out, not a requirement.

Your preparation checklist

Map the stated requirements to evidence
Prepare a one-page index of examples you can discuss for building and deploying speech, multimodal AI, or LLM systems; Python and PyTorch; infrastructure tooling; and communication with customers or partners. The posting asks for 5+ years of experience and a relevant degree or equivalent experience.
Rehearse an end-to-end project walkthrough
If you have a relevant project, prepare a concise account of its requirements, model selection, customization, deployment, observability, application integration, and your contribution. Include the latency, quality, and scale trade-offs you made; use actual results rather than invented figures.
Sketch a customer solution architecture
As a practice exercise, draw a proposed pipeline for streaming, multilingual transcription or conversational AI. Show where model choices, application integration, and cloud, data-center, or edge deployment decisions would sit; label the design as hypothetical.
Draft a speech validation plan
Prepare a sample test matrix for streaming recognition, multilingual configurations, and code-switching, which the role explicitly covers. Suggest how you would measure recognition quality, latency, and performance under load, and how findings would become technical feedback.
Review deployment and troubleshooting choices
Prepare a short comparison of how you would deploy and scale speech services using Docker, Kubernetes, Helm, or equivalent tools. Be ready to discuss where Riva, speech NIM microservices, or Nemotron might fit without claiming hands-on experience you do not have.
Practice a customer-facing technical review
Outline a brief demo or workshop that explains a voice AI pipeline to both technical partners and an account team. Include how you would handle a troubleshooting question and communicate a customer requirement to product or engineering teams.
Prepare questions about the role’s priorities
Bring questions about priority voice AI use cases in South East Asia, languages and code-switching needs, deployment environments, and how the team evaluates technical validations and partner enablement.

Availability last checked: 1 Oct 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…