New opportunity

Large Language Model Architect

Accenture · Hyderabad, India

About this role

Accenture seeks a Large Language Model Architect in Hyderabad for a full-time, hands-on AI LLM Technology Architecture role. The position designs, builds, integrates, tests and operationalizes enterprise-grade LLM, generative AI and agentic AI components for client engagements. It translates architecture into production-quality Databricks components for LLM applications, RAG pipelines, multi-agent workflows and AI platform integrations, contributing code, design decisions, reusable patterns and documentation as a technical or engineering lead.

Responsibilities

Build prompts, tools, agents, orchestration flows, memory and context handling, retrieval pipelines and evaluation harnesses. Implement ingestion, parsing, chunking, enrichment, embeddings, vector search and retrieval for structured and unstructured enterprise content.
Develop data-grounded lakehouse applications using Delta tables, Vector Search and governed features; use MLflow for tracing, evaluation and model lifecycle management, deploy through Model Serving and apply Unity Catalog governance.
Engineer PII detection and redaction, prompt-injection defenses, content filters, guardrails, authentication, authorization, lineage and audit logging. Work with architects, data engineers, product owners and security stakeholders to deliver tested, observable, maintainable components and produce designs, integration specifications, runbooks and evaluation results.

Requirements

Minimum 5 years of experience and 15 years of full-time education; a bachelor's degree in Computer Science, Computer Engineering, Data Science, AI/ML, Information Technology or a related engineering discipline is specified. The posting describes alternative experience profiles: typically 5+ years in software, data or AI engineering, including 2+ years in cloud-native engineering and 1+ year in GenAI, LLM, NLP or agentic AI delivery; or typically 7+ years, including 3+ years in cloud-native architecture or engineering and 1–2+ years in those AI areas.
Hands-on Python and Databricks experience, including Mosaic AI, Model Serving, Agent Framework, MLflow, Vector Search, Unity Catalog, Delta Lake, Lakehouse Monitoring, Feature Store, Workflows, Jobs, notebooks and Model Training. Understand RAG, tool calling, agent orchestration, prompt engineering, embeddings, evaluation, APIs, distributed systems, CI/CD, testing, observability and secure development. Experience delivering AI/ML or data products in financial services, healthcare, manufacturing, retail, telecom or life sciences is required.

Preferred

Databricks certifications; Spark/PySpark, Delta Live Tables, LangChain, LangGraph, LlamaIndex, Haystack, FastAPI, Docker or Kubernetes; model fine-tuning, Responsible AI, model risk management, synthetic data, human-in-the-loop review, A/B testing and lakehouse or GenAI cost optimization.

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

Skills for this role

Large Language Models (LLMs)Generative AIAgentic AIRAGPythonDatabricksDatabricks Mosaic AIDatabricks Model ServingDatabricks Agent FrameworkMLflowDatabricks Vector SearchUnity CatalogDelta LakeLakehouse MonitoringDatabricks Feature StoreDatabricks WorkflowsDatabricks JobsDatabricks Model TrainingPrompt EngineeringEmbeddingsVector DatabasesAPI DevelopmentDistributed SystemsCI/CDAutomated TestingObservabilitySecure SDLCGitInfrastructure as CodePII Detection and Redaction Engineering

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

Turn architecture into working components
The role is a hands-on technical or engineering lead expected to translate designs into production-quality LLM applications and integrations, contributing code, design decisions, reusable patterns and documentation.
Build data-grounded applications on Databricks
The posting emphasizes RAG and agentic applications using enterprise content, Delta tables, Vector Search and governed features, including ingestion through retrieval and integration with tools and agents.
Make systems testable and operable
Responsibilities include evaluation harnesses, MLflow tracing and evaluation, Model Serving deployments, observable components, runbooks and evaluation results—not just prototypes.
Engineer security and governance into the design
The role explicitly covers PII redaction, prompt-injection defenses, guardrails, access controls, lineage and audit logging, in collaboration with security and other stakeholders.

Likely challenges to think through

Balancing retrieval quality with operational cost
Preparation hypothesis, not a stated employer problem: richer parsing, chunking and retrieval may improve grounding but add latency and cost. Prepare to explain how you would evaluate answer quality alongside performance and cost in a Databricks RAG pipeline.
Controlling agents without losing usefulness
Preparation hypothesis: tool access and multi-agent orchestration can expand capability while increasing security and reliability risks. Prepare a design rationale for bounded permissions, prompt-injection defenses, tracing and tests.
Reusing patterns across different client domains
Preparation hypothesis: reusable components may need adaptation to each domain’s data, process constraints and controls. Prepare to distinguish what you would standardize from what you would validate with product, data and security stakeholders.

Practice your answers

Technical practice

How would you turn a high-level LLM architecture into a production Databricks application?

What to coverThe role calls for hands-on delivery of LLM applications and platform integrations on Databricks. Suggested preparation: explain component boundaries, data and model flows, deployment choices, and how you would make the design testable and maintainable.

How would you build retrieval over structured and unstructured enterprise content?

What to coverThe posting specifies ingestion, parsing, chunking, enrichment, embeddings, Delta tables, Vector Search and retrieval. Suggested preparation: discuss data-quality checks, chunking and indexing trade-offs, access to source content, and how you would diagnose poor retrieval.

How would you design a multi-agent workflow with tools and memory?

What to coverThe role includes agents, tool calling, orchestration flows and memory/context handling. Suggested preparation: explain when multiple agents are justified, how tool inputs and outputs are constrained, and how you would control context growth and failures.

How would you evaluate a data-grounded LLM application?

What to coverThe posting calls for evaluation harnesses, MLflow tracing and evaluation results. Suggested preparation: distinguish retrieval quality from answer quality, describe a representative test set and failure analysis, and explain how results would guide changes.

How would you deploy and manage an LLM component through its lifecycle?

What to coverThe role names MLflow model lifecycle management, Model Serving, Databricks Workflows and Jobs. Suggested preparation: walk through versioning, automated tests, deployment and rollback, making clear how you would trace a served response to its inputs and component versions.

How would you protect an LLM application against data leakage and prompt injection?

What to coverThe posting explicitly calls for PII detection and redaction, prompt-injection defenses, content filters, guardrails, authentication and authorization. Suggested preparation: place controls at ingestion, retrieval, tool use and output, and explain how you would test both legitimate access and attempted misuse.

How would you govern data and model use in a lakehouse-based AI system?

What to coverThe role specifies Unity Catalog governance, lineage, audit logging and governed features. Suggested preparation: describe how permissions and lineage would carry through source data, retrieval and serving, and how audit evidence would help investigate an output.

How would you monitor and improve a production LLM application?

What to coverThe posting calls for observability, Lakehouse Monitoring, reliability, scalability, performance and cost controls. Suggested preparation: identify useful quality and operational signals, explain how you would investigate degradation, and discuss trade-offs among latency, quality and cost.

How would you engineer and test the APIs and integrations around an LLM component?

What to coverThe role requires Python, APIs, distributed-systems knowledge, automated testing, CI/CD and secure development practices. Suggested preparation: discuss interface contracts, timeouts and retries, tests for external dependencies, and safeguards before a client-facing release.

When would you combine traditional ML components with a GenAI application?

What to coverThe posting includes feature and data preparation, Model Training, Feature Store, deployment and continuous improvement alongside GenAI. Suggested preparation: describe a plausible division of responsibilities between predictive components and an LLM, and how you would evaluate and version each without assuming a particular client use case.

Behavioural practice

Tell me about a time you led a technical design into working software.

What to coverThe role expects a hands-on technical or engineering lead who contributes code and design decisions. Suggested preparation: use a real situation, explain your architectural and implementation actions and trade-offs, and state the observable result.

Tell me about a time you resolved a disagreement across engineering, product and security stakeholders.

What to coverThe posting calls for collaboration with architects, data engineers, product owners and security stakeholders. Suggested preparation: describe the situation and competing concerns, what you did to reach a decision, and the resulting delivery or control outcome.

Tell me about a time industry constraints changed your AI or data-product approach.

What to coverThe role asks for practical industry experience to identify domain data, process constraints, controls and adoption risks. Suggested preparation: choose a real situation, explain the constraint and your response, and give the result without assuming any particular client domain.

Tell me about a production issue you diagnosed and addressed.

What to coverThe posting emphasizes tested, observable, maintainable components and continuous improvement. Suggested preparation: set out the symptoms, evidence and actions you took, then explain the result and any change that reduced recurrence.

Tell me about a reusable engineering pattern or technical artifact you created.

What to coverThe role includes reusable patterns, component designs, integration specifications, runbooks and evaluation results. Suggested preparation: describe the original need, your contribution and how others used it, then give a concrete outcome.

Sharpen your knowledge

Topics to prepare

LLM application architecture
The role requires RAG, tool calling, agent orchestration, prompts and context handling. Suggested revision: diagram a data-grounded application and explain when retrieval, a tool call or a multi-agent workflow is appropriate, including failure modes.
Databricks lakehouse architecture
The posting names Delta Lake, Vector Search, governed features, Unity Catalog, Mosaic AI and Model Serving. Suggested revision: map how content moves from Delta tables through retrieval to a served application, and explain where governance applies.
Evaluation and model lifecycle
The role calls for evaluation metrics, MLflow tracing and evaluation, model lifecycle management and continuous improvement. Suggested revision: outline how you would compare retrieval quality and answer quality, inspect traces and decide whether a change is ready to deploy.
Security, privacy and governance
Required controls include PII redaction, prompt-injection defenses, content filters, authentication, authorization, lineage and audit logging. Suggested revision: threat-model a RAG workflow and explain which control addresses each identified risk.
Production engineering tradeoffs
The posting requires cloud-native engineering knowledge, APIs, distributed systems, testing, observability, reliability, scalability and cost controls. Suggested revision: review an application design and explain its likely failure points, monitoring signals and performance-versus-cost tradeoffs.
Industry constraints
Experience delivering AI/ML or data products in at least one listed industry is required. Suggested revision: choose a domain you know and outline its data, process, control and adoption constraints for a hypothetical GenAI use case; this is preparation, not a stated interview exercise.

Skills to revise

Build a Python retrieval pipeline
The role requires Python and ingestion, parsing, chunking, embeddings, vector search and retrieval for enterprise content. Suggested exercise: implement a small pipeline over sample structured and unstructured content, then test whether retrieved passages support the answers.
Implement an agent workflow
The posting calls for prompts, tools, agents, orchestration and memory/context handling, with Databricks Agent Framework experience. Suggested exercise: build a small tool-calling flow and test tool selection, context handling and error cases.
Connect Databricks components
Hands-on experience is requested across Delta Lake, Vector Search, Unity Catalog, Mosaic AI, Model Serving, Feature Store, Workflows, Jobs, notebooks and Model Training. Suggested exercise: using a practice environment if available, assemble a small governed data-to-serving workflow and document what each component does.
Evaluate and observe a deployed component
MLflow tracing/evaluation, Lakehouse Monitoring, automated testing and observability are named in the posting. Suggested exercise: create a small evaluation set, add regression tests, inspect traces and write down a monitoring check for a changed prompt or retrieval step.
Harden and document an integration
The role includes secure development, CI/CD, Git, infrastructure-as-code and artifacts such as integration specifications and runbooks. Suggested exercise: add PII redaction and a prompt-injection test to a sample flow, then draft a concise integration specification and deployment runbook.

Your preparation checklist

Map the required experience to evidence
The posting specifies Python, Databricks, cloud-native engineering, GenAI delivery and AI/ML or data-product delivery in a listed industry. Prepare a one-page evidence map with specific examples and dates; leave gaps explicit rather than implying experience.
Prepare a production project walkthrough
Choose a real LLM, GenAI or data-product project to explain your architecture decisions, code contributions, integration points, testing and operational results. Identify what you personally owned and how domain constraints affected the design.
Sketch a Databricks RAG design
Prepare a labeled design showing ingestion, parsing, chunking, embeddings, Delta tables, Vector Search and retrieval for enterprise content. Explain where Unity Catalog governance, Model Serving and monitoring fit, and state any assumptions.
Rehearse an agent and evaluation exercise
Sketch a tool-calling workflow with orchestration and memory or context handling. Prepare example evaluation cases, metrics and failure analysis, including how MLflow tracing and evaluation would help diagnose results.
Show how you would engineer safety controls
Prepare a threat-and-controls table covering PII detection and redaction, prompt injection, content filters, authentication, authorization, lineage and audit logging. Explain where each control runs and how you would test it.
Assemble engineering-lead artifacts
Bring an anonymized or illustrative component design, integration specification and deployment runbook. Be ready to discuss Python APIs, Git-based development, automated tests, CI/CD, observability and trade-offs in reliability, scale and cost.
Write questions about delivery expectations
Ask which Databricks capabilities the team uses in current engagements, how architects and security stakeholders divide decisions, and what evaluation results and operational artifacts are expected from this lead role.

Availability last checked: 5 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…