New opportunity

Data Scientist, Specialist

Vanguard · Hyderabad, India

About this role

Vanguard seeks a Data Scientist, Specialist for its Hyderabad, Telangana office. The role leads the design, governance, and deployment of enterprise AI agents that combine large language models, enterprise data, and agentic workflows to automate knowledge-intensive processes. It partners with business, risk, compliance, and technology teams to deliver measurable outcomes. The position follows a hybrid work model; only qualified external applicants will be considered.

Responsibilities

Build enterprise knowledge systems that integrate structured and unstructured data so AI agents can retrieve, reason over, and use trusted information.
Identify and prioritize business problems suited to agentic AI, LLMs, and NLP. Design agents that reason, plan, and act across tools, workflows, and data sources, including multi-agent and tool-augmented solutions for complex processes.
Establish business-focused evaluation covering relevance, coverage, trustworthiness, explainability, and adoption alongside model performance. Work with engineers on scalable, secure, performant deployment and implement monitoring and guardrails for bias, drift, and failures.
Align solutions with risk, compliance, and responsible AI standards. Advise teams on appropriate uses of AI, contribute reusable practices and strategic direction, and mentor colleagues.

Qualifications

Minimum 8 years in data science, predictive analytics, or advanced statistical modeling roles; 3+ years delivering AI/ML solutions in production; and 5+ years of hands-on Python experience.
Experience designing agents and building enterprise LLM applications, including RAG, tool use, orchestration, and evaluation. Strong NLP experience with unstructured text and language-driven workflows, business problem-solving ability, communication skills, and experience in cross-functional enterprise settings.
The posting specifies a bachelor's degree in Analytics, Applied Mathematics, Economics, Statistics, or Computer Science, or a master's degree or diploma in a related quantitative field; it also lists an MS or PhD in Computer Science, Machine Learning, Data Science, or a related quantitative field. Distributed data processing experience is a plus.

Vanguard describes financial, physical, and personal wellness support among its benefits.

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

Skills for this role

Agentic AIAI agentsLarge Language ModelsRetrieval-Augmented GenerationNatural Language ProcessingPythonData sciencePredictive analyticsStatistical modelingDistributed data processingLLM orchestrationLLM evaluationAI monitoringResponsible AIRisk managementStakeholder collaborationCommunicationMentoring

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

Choose AI use cases by business value
The role partners with business stakeholders to frame and prioritize problems suited to agentic AI, LLMs, and NLP, with measurable outcomes rather than model performance alone.
Build trusted enterprise knowledge systems
The posting calls for integrating structured and unstructured data across business platforms so agents can retrieve, reason over, and use reliable organizational knowledge.
Design agents for multi-step work
A central duty is to architect agents that reason, plan, and act across tools, workflows, and data sources, including multi-agent and tool-augmented solutions.
Evaluate and govern production systems
The role defines business-focused measures such as coverage, relevance, trustworthiness, explainability, and adoption; works with engineers on secure, scalable deployment; and implements monitoring and guardrails for bias, drift, and failures.
Lead across functions
The specialist advises on where AI does and does not add value, aligns with risk and compliance stakeholders, contributes reusable practices, and mentors colleagues.

Likely challenges to think through

Useful retrieval versus trusted information
Preparation hypothesis: combining data from multiple platforms may make broad coverage difficult to reconcile with reliable answers. Prepare to explain how you would test retrieval quality, identify unsupported outputs, and limit an agent's use of uncertain information.
Agent autonomy versus control
Preparation hypothesis: agents that act across tools can automate more steps but create more consequential failure modes. Prepare a design rationale for action boundaries, guardrails, monitoring, and escalation that preserves useful automation.
Business adoption versus technical scores
Preparation hypothesis: a system can perform well on model tests without improving a workflow. Prepare to connect evaluation of relevance and explainability to coverage, adoption, and measurable business outcomes, and to explain when a simpler non-agentic approach may be preferable.

Practice your answers

Technical practice

How would you decide whether a business process warrants an AI agent rather than a simpler solution?

What to coverSuggested preparation: Describe how you would assess process complexity, available data, required actions, risks, and the value of automation. Explain how you would prioritize measurable business outcomes over model performance alone.

How would you build an enterprise knowledge system from structured and unstructured data?

What to coverSuggested preparation: Cover ingestion, data quality, access controls, and how an agent would trace retrieved information to a trusted source. Discuss how you would handle conflicting or outdated information across business platforms.

How would you design and evaluate retrieval-augmented generation for a language-driven workflow?

What to coverSuggested preparation: Explain choices for preparing unstructured text, retrieving relevant material, and grounding responses. Discuss the trade-off between broad coverage and precise, trustworthy answers.

How would you design an agent that plans and takes actions across tools and workflows?

What to coverSuggested preparation: Walk through tool selection, action boundaries, and checks before consequential actions. Explain how the design would handle failed tool calls or insufficient information.

When would you use a multi-agent design instead of a single tool-augmented agent?

What to coverSuggested preparation: Compare separation of responsibilities with added orchestration complexity, latency, and failure points. Describe how you would test whether multiple agents improve a complex, multi-step process.

What would a business-focused evaluation framework for an enterprise AI agent include?

What to coverSuggested preparation: Address coverage, relevance, trustworthiness, explainability, and user adoption alongside technical model performance, as specified in the posting. Explain how you would connect those measures to a defined business outcome.

How would you investigate an agent that starts producing less reliable results after deployment?

What to coverSuggested preparation: Describe how monitoring could distinguish data drift, retrieval problems, tool failures, and other failure modes. Discuss how you would evaluate the impact and decide whether to restrict, correct, or roll back the system.

How would you work with engineers to deploy an agent securely and at scale?

What to coverSuggested preparation: Discuss access to enterprise data, performance, reliability, and the operational trade-offs of tool-augmented workflows. Explain which behaviors you would validate before production and continue monitoring afterward.

How would you assess bias and align an AI agent with risk, compliance, and responsible AI standards?

What to coverSuggested preparation: Outline tests and guardrails for problematic outputs or actions, including how findings would be documented for review. Explain how governance requirements could change a proposed agent design.

How would you use diagnostic, predictive, or prescriptive analysis to support an AI use case?

What to coverSuggested preparation: The role's page summary names these forms of analysis, while its detailed description emphasizes business-driven AI. As a suggested approach, explain how you would establish a baseline, assess uncertainty, and determine whether an agent improved the decision or process.

Behavioural practice

Tell me about a time you aligned business and technical stakeholders on an AI problem.

What to coverSuggested preparation: Use a real situation, the actions you took to clarify priorities and constraints, and the result. Explain how you communicated the proposed value in terms both groups could assess.

Describe a time you advised against using AI for a proposed problem.

What to coverSuggested preparation: Give the situation, how you evaluated alternatives and explained your recommendation, and the result. Focus on the posting's expectation to advise teams where agentic AI and LLMs do not add value.

Tell me about a time you resolved a risk or compliance concern in a cross-functional project.

What to coverSuggested preparation: State the concern, your actions with relevant stakeholders, and the outcome. Distinguish the safeguard you put in place from any trade-off it created for delivery or usability.

Describe a time you mentored a colleague on an applied AI or data science problem.

What to coverSuggested preparation: Explain the colleague's challenge, what you did to develop their judgment or approach, and the result. Keep the example grounded in your own actions rather than claiming the team's work as yours.

Tell me about a production AI or analytics solution whose business impact you evaluated.

What to coverSuggested preparation: Describe the situation, how you defined and measured the intended outcome, and what the results showed. Include how you responded to adoption or performance evidence that challenged the initial plan.

Sharpen your knowledge

Topics to prepare

Choosing business problems for agentic AI
Posting: The role frames and prioritizes high-value problems with business stakeholders and focuses on measurable outcomes. Practice: Compare two hypothetical knowledge-intensive processes, explain when an agent would or would not add value, and define an outcome measure for each.
Enterprise knowledge and NLP
Posting: Agents must use trusted knowledge from structured and unstructured data, and the role calls for strong NLP experience. Practice: Sketch how an agent would retrieve information from both types of data, then explain how it would handle conflicting or incomplete evidence.
Agent architecture and orchestration
Posting: The role designs agents that reason, plan, and act across tools and workflows, including multi-agent solutions. Practice: Diagram a multi-step workflow and justify where tool use, agent handoffs, and checks before actions belong.
Business-focused LLM evaluation
Posting: Evaluation must cover relevance, coverage, trustworthiness, explainability, and adoption alongside model performance. Practice: Build a sample evaluation matrix with a test question, a measure, and a failure criterion for each dimension.
Responsible production AI
Posting: The role addresses scalable, secure deployment, monitoring, guardrails, bias, drift, failures, risk, and compliance. Practice: Review a hypothetical deployed agent failure and outline what should be detected, contained, investigated, and communicated.
Statistical and analytical reasoning
Posting: The qualifications call for substantial data science, predictive analytics, or advanced statistical modeling experience; the role also supports business decision-making. Practice: Explain how you would distinguish a genuine improvement in a business outcome from sampling variation or a change in who uses the system.

Skills to revise

Python implementation
Posting: The role specifies hands-on Python experience. Practice: Write a small Python workflow that accepts a question, processes retrieved text, and returns an answer with its supporting passages; test empty and contradictory results.
RAG over mixed enterprise data
Posting: Enterprise LLM application experience includes RAG, and the role integrates structured and unstructured information. Practice: Using sample data, assemble a retrieval-and-answer flow and inspect whether each answer is supported by the retrieved records.
Tool-augmented workflow design
Posting: The role builds agents that act across tools and data sources. Practice: Prototype a multi-step agent flow with a mock tool, including input checks, a failed-tool response, and a record of actions taken.
Evaluation and monitoring artifacts
Posting: The role implements evaluation, monitoring, and guardrails for LLM and agentic systems. Practice: Create a small test set of normal and failure cases, score the outputs against stated criteria, and specify what changes would trigger review after deployment.
Cross-functional explanation and mentoring
Posting: The role partners with business, risk, compliance, and technology teams, advises on appropriate AI use, and mentors colleagues. Practice: Prepare a brief design explanation for a non-technical stakeholder and a technical peer, covering expected value, limitations, and governance choices.

Your preparation checklist

Map evidence to the qualifications
Prepare a one-page matrix of examples and dates for the stated data science, production AI/ML, and Python experience requirements. Add evidence of NLP, enterprise LLM applications, and cross-functional work; leave gaps blank rather than inventing examples.
Frame a business problem for agentic AI
Draft a short decision brief comparing a knowledge-intensive process suitable for an agent with a simpler alternative. State the expected business outcome, risks, and how you would prioritize the work; this is a preparation exercise, not a stated Vanguard project.
Prepare a project walkthrough
If you have a relevant project, outline the problem, your decisions, structured and unstructured data, retrieval or NLP approach, production responsibilities, and measured outcome. Be ready to explain trade-offs and what you would change.
Sketch an enterprise agent design
For the hypothetical process, draw a simple architecture showing trusted knowledge sources, retrieval, tool use, orchestration, and human oversight. Explain when a single agent would suffice and when a multi-agent approach would be justified.
Build an evaluation scorecard
Draft measures for coverage, relevance, trustworthiness, explainability, and adoption alongside technical performance. Include example test cases and a way to connect results to the business outcome.
Plan deployment and controls
Prepare a deployment checklist covering scalability, security, monitoring, and responses to bias, drift, and agent failures. Note decisions you would review with engineering, risk, and compliance rather than assuming Vanguard’s specific procedures.
Rehearse senior-level collaboration examples
Prepare concise examples, where available, of advising stakeholders when not to use AI, explaining technical trade-offs to non-technical audiences, establishing reusable practices, and mentoring colleagues.
Write questions for the employer
Bring questions about priority knowledge workflows, available data and tools, success measures, ownership of agent evaluations and guardrails, and how business, engineering, risk, and compliance teams collaborate.

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…