New opportunity

Specialist Solutions Architect - AI/ML

Databricks · Bengaluru, India

About this role

Databricks seeks an AI/ML Specialist Solutions Architect to serve as a technical expert for customers and its Field Engineering organization. Working with Solution Architects, this person will help customers design production-grade ML and AI applications on the Databricks Data Intelligence Platform and align technical roadmaps with the platform’s development. The role is hybrid and based in Bengaluru, India.

Responsibilities

Architect, build, scale and optimize customer AI/ML workloads, including agents, end-to-end ML pipelines, training and inference optimization, cloud-native integrations and production MLOps.
Advise on enterprise generative AI solutions, including RAG, tool-calling and multi-agent systems, guardrails, natural-language querying of structured data, evaluation, observability and monitoring.
Support Solution Architects during technical sales across feature engineering, training, tracking, serving and model monitoring. Share expertise within the ML specialist community and mentor others.
Work with product and engineering teams to represent customer needs, influence priorities and the product roadmap, and support adoption of Databricks AI offerings.

Requirements

At least 5 years of hands-on industry ML experience, either building and maintaining production-grade cloud infrastructure on AWS, Azure or GCP for ML application deployment, including drift monitoring; or working with LLMs and agentic systems, including vector databases, LLM fine-tuning, AI guardrails and deployment using tools such as Hugging Face, LangChain and OpenAI.
A graduate degree in a quantitative discipline, such as computer science, engineering, statistics or operations research, or equivalent practical experience. Ability to explain or teach technical concepts to technical and nontechnical audiences, collaborate, keep learning and deliver business value through ML and AI.
Ability to meet technical training and role-specific expectations within three months of hire and travel up to 30% when needed.

Preferred

At least 2 years of customer-facing pre-sales or post-sales experience. Databricks notes that benefits vary by region but does not specify them for this role.

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

Skills for this role

Machine LearningGenerative AIMLOpsLLMsRAGAI agentsMulti-agent orchestrationAI guardrailsAI evaluationAI observabilityModel monitoringDrift monitoringFeature engineeringModel trainingModel servingInference optimizationML pipelinesVector databasesLLM fine-tuningHugging FaceLangChainOpenAIAWSAzureGoogle Cloud PlatformCloud infrastructureSolutions architectureTechnical pre-salesTechnical 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

Production-grade AI/ML architecture
The role architects, builds, scales and optimizes customer workloads on Databricks, including agents, ML pipelines, cloud-native integrations, training and inference optimization, and MLOps. Prepare to explain architectural choices across deployment, reliability and monitoring.
Enterprise generative AI design
The posting explicitly covers RAG, tool-calling and multi-agent systems, guardrails, natural-language querying of structured data, evaluation and observability. Prepare to explain how these components fit into a production design and how their performance would be evaluated.
Technical guidance during customer sales
The specialist supports Solution Architects during technical sales, from feature engineering and training through tracking, serving and model monitoring. Prepare to communicate design decisions and business value to both technical and nontechnical audiences.
Customer feedback and technical leadership
The role represents customer needs to product and engineering teams, influences priorities, participates in the ML specialist community and mentors others. Prepare examples of translating technical feedback into clear priorities and sharing expertise.

Likely challenges to think through

Balancing capability with production controls
Likely preparation hypothesis: adding agents or RAG can expand what an application can do while making evaluation, guardrails and observability harder. Prepare a design walkthrough that states the trade-offs and explains how you would test and monitor the system.
Unifying an end-to-end ML workflow
Likely preparation hypothesis: optimizing training or inference alone may not solve problems elsewhere in a pipeline. Prepare to trace a workload from features and training through tracking, serving and drift monitoring, explaining where you would investigate bottlenecks.
Reconciling customer needs with an evolving platform
Likely preparation hypothesis: customers may want a specific technical roadmap while the Databricks platform continues to change. Prepare to explain how you would distinguish immediate architectural needs from product feedback and communicate that distinction to Solution Architects and product teams.

Practice your answers

Technical practice

How would you design an end-to-end ML pipeline for a production customer workload?

What to coverSuggested preparation: Walk through feature engineering, training, experiment tracking, deployment and model monitoring, showing how each stage supports reproducibility. Explain what you would clarify about the customer's workload before choosing an architecture.

How would you integrate an ML application with cloud-native services?

What to coverSuggested preparation: Describe the infrastructure and integration decisions you would make for a production deployment on AWS, Azure or GCP. Discuss reliability, scaling and operational trade-offs without assuming a particular customer cloud.

How would you diagnose and improve slow model training or inference?

What to coverSuggested preparation: Explain how you would measure where time and resources are spent before changing the workload. Compare possible optimizations against their effects on cost, latency and model quality.

How would you detect and respond to drift in a deployed ML application?

What to coverSuggested preparation: Distinguish changes in input data from changes in model performance and explain what evidence you would monitor. Describe how you would investigate an alert and decide whether intervention is warranted.

How would you design and evaluate a retrieval-augmented generation application?

What to coverSuggested preparation: Cover retrieval, the role of a vector database and how retrieved material reaches the model. Explain how you would test retrieval quality separately from answer quality and weigh those results against latency.

When would you use a tool-calling agent rather than multi-agent orchestration?

What to coverSuggested preparation: Compare the tasks each approach can handle with its added complexity and operational cost. Explain how you would inspect tool use and failure paths before recommending a design.

How would you choose guardrails for an enterprise GenAI application?

What to coverSuggested preparation: Identify the risks the application needs to control and where checks could operate in the workflow. Discuss how you would test whether a guardrail reduces harmful behavior without blocking legitimate use.

How would you assess natural-language querying of structured data?

What to coverSuggested preparation: Explain how you would check that generated queries reflect user intent and return accurate results. Address how you would handle ambiguity and limit access to data the user should not retrieve.

What would you monitor after deploying a GenAI workload?

What to coverSuggested preparation: Separate pre-deployment evaluation from production observability, and identify signals for answer quality, failures and workload performance. Explain how those signals would help you troubleshoot a regression.

When would you consider fine-tuning an LLM?

What to coverSuggested preparation: Explain what evidence would justify fine-tuning rather than changing prompts or retrieval. Discuss the data, evaluation and deployment considerations you would review before recommending it.

Behavioural practice

Tell me about a time you explained a complex ML decision to technical and nontechnical stakeholders.

What to coverSuggested preparation: Give the situation, the decision each audience needed to make and how you adapted your explanation. Describe the resulting decision or understanding rather than relying on technical detail alone.

Describe a time you helped a customer or partner choose between competing AI/ML approaches.

What to coverSuggested preparation: Set out the customer's objective, your role and the trade-offs you made explicit. Explain what action followed and what evidence showed whether the choice delivered value.

Tell me about a time you mentored others on an ML or AI topic.

What to coverSuggested preparation: Describe the knowledge gap, how you helped and how you adjusted to the other person's experience. Explain the observable result for their work or understanding.

Describe a time you brought customer feedback to a product or engineering team.

What to coverSuggested preparation: Explain the customer need, how you validated and communicated it, and any competing priorities you encountered. State what happened next without claiming influence you cannot substantiate.

Tell me about a time you had to learn a new AI/ML technique quickly to deliver an outcome.

What to coverSuggested preparation: Describe the situation, how you tested your understanding and how you applied what you learned. Explain the result and any limitation or follow-up work you identified.

Sharpen your knowledge

Topics to prepare

Production ML architecture
The posting calls for production-grade workloads spanning feature engineering, training, tracking, serving, monitoring, and cloud-native integration. Revision exercise: draw an end-to-end architecture on the Databricks platform and explain where each stage operates and how failures would be detected.
Enterprise RAG and agent design
The posting names RAG, vector databases, tool-calling agents, multi-agent orchestration, and natural-language querying of structured data. Revision exercise: compare two designs for a customer question-answering application, explaining when retrieval, tools, or structured-data queries fit and what each adds to production complexity.
AI evaluation, guardrails, and observability
The role explicitly covers guardrails, evaluation, observability, and monitoring. Revision exercise: outline how you would assess answer quality and unsafe behavior before release, then detect regressions after deployment.
Training and inference tradeoffs
The posting calls for training and inference optimization and cites LLM fine-tuning within the AI Engineer experience path. Revision exercise: explain when you would improve retrieval, fine-tune a model, or optimize inference, stating the quality, latency, and operational tradeoffs.
Customer-facing technical judgment
The role involves advising customers, supporting Solution Architects during technical sales, and conveying customer needs to product and engineering teams. Revision exercise: practice explaining one architecture decision to both a technical team and a nontechnical stakeholder, then summarize the resulting product feedback. This exercise is a preparation suggestion, not a stated interview format.

Skills to revise

Build an end-to-end ML workflow
Suggested practice: assemble a small workflow covering feature preparation, training, experiment tracking, serving, and model monitoring; rehearse diagnosing a failed stage. These activities reflect the posting’s technical-sales support scope, not a requirement to use any particular implementation.
Plan cloud deployment and drift monitoring
Suggested practice for the posting’s ML Engineer experience path: sketch a production ML deployment on one of AWS, Azure, or GCP, identify operational failure points, and define how you would detect and respond to drift. The posting presents this as one of two alternative hands-on experience paths.
Prototype a RAG or tool-calling workflow
Suggested practice for the posting’s AI Engineer experience path: build a small retrieval workflow with a vector database or a tool-calling agent, then test what happens when retrieval is irrelevant or a tool fails. Both patterns are explicitly named in the role.
Test guardrails and monitor an AI application
Suggested practice: create a compact set of normal and adversarial prompts, record evaluation results, and specify signals to monitor in production. This applies the posting’s guardrails, evaluation, observability, and monitoring responsibilities.
Walk through a customer architecture decision
Suggested practice: deliver a short design walkthrough for a production AI workload, explain tradeoffs to a Solution Architect and a nontechnical customer audience, and translate one customer constraint into actionable product feedback. This mirrors the role’s communication and cross-functional responsibilities; the walkthrough is a preparation suggestion.

Your preparation checklist

Map evidence to one experience path
The posting asks for 5+ years of hands-on ML experience in either production cloud ML infrastructure or LLM and agentic systems. Prepare a concise timeline with specific work you can substantiate for the applicable path; note your degree or equivalent practical experience separately.
Build a production project walkthrough
Choose a real project, if available, and prepare a one-page architecture diagram. Explain your decisions, your contribution, production constraints, measurable results and what you would change. Connect it to the role’s focus on building and scaling customer AI/ML workloads without implying it used Databricks if it did not.
Practice an enterprise GenAI design exercise
As interview preparation, sketch a RAG or tool-calling agent design for a hypothetical customer. Be ready to justify retrieval, guardrails, evaluation, observability and monitoring; state assumptions rather than presenting the scenario as a Databricks project.
Trace a model from training to monitoring
Prepare a short technical walkthrough of feature engineering, training, tracking, serving and model monitoring. Include cloud-native integration, drift detection and a training or inference optimization trade-off where your experience supports it.
Rehearse a customer-facing explanation
The role supports Solution Architects during technical sales and calls for communication with technical and nontechnical audiences. Practice explaining one architecture twice: once as technical design choices and once as business value, risks and trade-offs.
Prepare a customer-feedback example
Identify a real instance, if available, where you translated user needs into a technical priority or collaborated with product or engineering. Explain the evidence behind the request and the outcome; this prepares for the posted responsibility to represent customers and influence the roadmap.
Write focused questions for the employer
Prepare questions about how this specialist partners with Solution Architects, what technical training and outcomes are expected in the first three months, and how customer feedback reaches product teams. Confirm how hybrid work in Bengaluru and travel of up to 30% operate for this role.

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 India. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…