New opportunity

Ai Engineer

Tristone Partners LLP · Mumbai, India

About this role

Tristone Partners LLP is hiring an Ai Engineer in Mumbai, India, to help embed agentic AI into its workflows and systems. The role focuses on building agents, MCP servers and evaluation loops that make AI-generated code reliable, auditable and efficient to deliver. The engineer will work with senior team members on AI integrations.

Responsibilities

Implement agents, MCP servers and supporting services that propose and apply changes to large codebases.
Help develop retrieval systems that provide agents and developers with current context from codebases and design artifacts.
Build compile, test and evaluation pipelines for AI-generated changes, including static analysis, style and safety checks, performance gates and code review.
Apply practices for concurrency, telemetry, configuration hygiene and performance-sensitive code; collaborate on experiments to improve AI-driven workflows.
Follow IT policies and maintain information security.

Requirements

At least 1 year of experience; proficiency in a high-level programming language, preferably Java or Python; experience contributing to developer-facing or production systems; interest in learning AI and LLM application patterns; and strong communication skills. A BS or higher in Computer Science or a related field is required, along with strong fundamentals in algorithms, data structures and systems.

Strongly preferred: Exposure to AI tools or workflows such as LangChain, LangGraph or AutoGen; retrieval methods including vector search, embeddings or hybrid approaches; compiler or static-analysis tools or large-scale refactoring; interest in customizing or fine-tuning open-weight models; and knowledge of model-serving or evaluation infrastructure.

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

Skills for this role

AI agentsMCP serversRetrieval systemsVector searchEmbeddingsStatic analysisModel evaluationModel servingFine-tuningConcurrencyTelemetryJavaPythonLangChainLangGraphAutoGenCode reviewCommunication

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

Build code-changing agents
Implement agents, MCP servers and supporting services that propose and apply changes to large codebases.
Provide current codebase context
Help develop retrieval systems that give agents and developers accurate, up-to-date context from codebases and design artifacts.
Validate AI-generated changes
Build compile, test and evaluation pipelines with static analysis, style and safety checks, performance gates and code review so changes are reliable and auditable.
Deliver dependable integrations
Apply practices for concurrency, telemetry, configuration hygiene and performance-sensitive code; collaborate with colleagues on workflow experiments while following IT and information-security policies.

Likely challenges to think through

Autonomy versus reviewability
Preparation hypothesis: applying agent-proposed changes across a large codebase may require balancing useful automation with changes engineers can inspect and audit. Prepare to explain how you would scope a change and make its proposed edits and validation results reviewable.
Fresh context versus retrieval cost
Preparation hypothesis: retrieving enough current code and design context may compete with latency and resource use. Prepare to discuss how you would check context relevance and freshness, and measure whether retrieval improves outcomes.
Validation depth versus delivery speed
Preparation hypothesis: compile, test, safety and performance gates can catch problems but slow iteration. Prepare a staged validation approach and explain what evidence you would require before an AI-generated change proceeds.

Practice your answers

Technical practice

How would you design an agent that proposes and applies a change to a large codebase?

What to coverSuggested preparation: Explain how the agent would identify relevant files, produce a reviewable change, and avoid applying an unverified edit. Discuss trade-offs between automation and human review; this is a design question, not a stated implementation requirement.

What would an MCP server need to expose to support an agent working with a codebase?

What to coverSuggested preparation: Describe a narrow interface for accessing context or proposing changes, with attention to permissions and observable results. Relate the design to the posting’s work on MCP servers and information security.

How would you keep retrieved code and design-artifact context accurate and up to date?

What to coverSuggested preparation: Discuss indexing, refresh or invalidation, and how an agent could distinguish current context from stale results. Explain how you would check whether retrieved material actually supports a proposed code change.

When might vector search, embeddings or a hybrid retrieval approach help with codebase context?

What to coverSuggested preparation: Compare semantic matching with exact matches for identifiers, paths and symbols. These retrieval methods are strongly preferred exposure in the posting, so focus on reasoned trade-offs rather than assuming a particular system is in use.

What checks would you put between an AI-generated code change and its acceptance?

What to coverSuggested preparation: Outline how compilation, tests, style and safety checks, performance gates, and code review can catch different failures. Explain what you would do when checks disagree or a change passes tests but remains difficult to review.

How would you evaluate whether an AI-driven code-change workflow is improving?

What to coverSuggested preparation: Propose repeatable tasks and measures such as change correctness, review effort and time to a validated result. Discuss how you would inspect failures rather than relying on a single aggregate score.

How could static analysis help assess or carry out a large-scale code change?

What to coverSuggested preparation: Explain what structural checks might detect that compilation or tests miss, and where static analysis can produce false positives. Familiarity with static-analysis or refactoring tools is strongly preferred, not listed as a required qualification.

What concurrency risks arise when supporting services handle multiple code-change tasks?

What to coverSuggested preparation: Discuss shared state, conflicting edits and resource contention, then describe ways to preserve correctness. Connect the answer to the role’s stated emphasis on concurrency and reliable outputs.

How would telemetry and configuration hygiene help troubleshoot a slow or unreliable AI integration?

What to coverSuggested preparation: Describe what you would measure across retrieval, generation and validation, while avoiding exposure of sensitive information. Explain how controlled configuration and performance measurements help isolate a regression.

What factors would you examine when choosing between model-serving approaches or customizing an open-weight model?

What to coverSuggested preparation: Compare operational complexity, latency, evaluation evidence and the value of customization against a simpler baseline. Model-serving knowledge and interest in fine-tuning are strongly preferred, so treat this as a trade-off discussion rather than an assumed project.

Behavioural practice

Tell me about a time you collaborated with a more experienced engineer on a technical change.

What to coverSuggested preparation: Use a real situation, your actions and the result to show how you sought input and incorporated feedback. This connects to the role’s work with senior team members without assuming prior AI experience.

Describe an experiment you used to decide whether a developer workflow should change.

What to coverSuggested preparation: Give the situation, what you compared, your contribution and the observed result. Be clear about inconclusive evidence or trade-offs, since the role involves collaborative experiments.

Tell me about a time you communicated a technical concern during a code review or shared change.

What to coverSuggested preparation: Explain the situation, how you made the concern actionable and what resulted. Choose an example that shows clear communication with engineers rather than claiming a particular review process at Tristone.

Describe a time you protected sensitive information or followed an IT policy while delivering work.

What to coverSuggested preparation: State the situation, the actions you took and the result without disclosing confidential details. Relate the example to the posting’s explicit information-security and IT-policy responsibilities.

Tell me about a time you learned an unfamiliar tool or technical pattern for a project.

What to coverSuggested preparation: Cover the situation, how you learned and tested the approach, and the result. The posting values curiosity about AI and LLM application patterns even without extensive prior exposure.

Sharpen your knowledge

Topics to prepare

Agent and MCP server design
Posting: The role implements agents, MCP servers and services that propose and apply changes to large codebases. Suggested revision exercise: diagram how an agent obtains code context, proposes a change and passes it to a service for application; identify where each step could fail.
Retrieval and context freshness
Posting: Retrieval must provide agents and developers with accurate, up-to-date context from codebases and design artifacts; vector search, embeddings and hybrid approaches are preferred exposure. Suggested revision exercise: compare those retrieval approaches for finding a recently changed function and its design rationale.
Evaluation of AI-generated code
Posting: The role helps build compile, test and evaluation pipelines with static-analysis, style, safety, performance and code-review checks. Suggested revision exercise: order these checks for a proposed patch and define what evidence would make the patch pass or fail each gate.
Systems and reliability fundamentals
Posting: Algorithms, data structures and systems fundamentals are qualifications; concurrency, telemetry, configuration hygiene and performance-sensitive paths appear in the responsibilities. Suggested revision exercise: analyze the time cost and potential race conditions of two agents updating context while a codebase changes.
Preferred AI workflow exposure
Posting: Exposure to LangChain, LangGraph or AutoGen, and knowledge of model-serving or evaluation infrastructure, are strongly preferred rather than required. Suggested revision exercise: explain how one named workflow tool could coordinate a retrieval-and-evaluation loop, and distinguish that from serving a model.

Skills to revise

Prototype a bounded code-change flow
Posting: Agents and MCP servers are used to propose and apply codebase changes. Suggested practice: in Java or Python, build a small MCP-backed flow that reads a sample repository, produces a proposed change and keeps application as a distinct step. This is a practice exercise, not a stated architecture requirement.
Refresh a retrieval index
Posting: Retrieval should reflect current code and design artifacts. Suggested practice: index a small set of code and design files, edit one file, refresh the index and check whether queries return the updated context. Vector or hybrid retrieval is preferred exposure, not required experience.
Run a patch through quality gates
Posting: Compile, test, static-analysis, style, safety, performance and review checks are named responsibilities. Suggested practice: make a small Java or Python change, run applicable checks, record failures and write a brief review explaining whether the change is ready to apply.
Instrument a concurrent service
Posting: The role applies concurrency, telemetry, configuration hygiene and performance practices. Suggested practice: run concurrent requests against a small service, record latency and errors, then change one configuration setting and explain its effect. The service is a preparation exercise, not a stated employer project.
Report an evaluation experiment
Posting: The engineer collaborates on experiments to improve AI-driven workflows and needs strong communication skills. Suggested practice: compare two versions of a code-change workflow on the same small task set, record check results and summarize the trade-offs for another engineer.

Your preparation checklist

Prepare a systems contribution walkthrough
Choose a developer-facing or production system you contributed to. Be ready to explain your Java, Python, or other high-level language work, the design trade-offs, your contribution, and how you communicated with engineers.
Sketch an agent and MCP service
As a preparation exercise, draw how an agent could propose and apply changes to a large codebase through an MCP server. Identify what would require review and how changes would remain auditable.
Design a codebase retrieval approach
Prepare a one-page design for giving agents current context from code and design artifacts. Explain how you would update the index and assess whether retrieved context is accurate; discuss vector, embedding, or hybrid methods if familiar.
Map an AI-code evaluation loop
Outline a sample path from proposed change to code review, including compile, test, static-analysis, style, safety, and performance checks. State what each check would catch and what evidence you would record.
Review reliability and systems fundamentals
Prepare a concrete example or small design exercise covering concurrency, telemetry, configuration hygiene, or performance-sensitive code. Be ready to explain the relevant algorithms, data structures, and systems trade-offs.
Distinguish experience from learning plans
List any direct exposure to agent frameworks, model serving, evaluation infrastructure, or open-weight models separately from topics you would need to learn. The posting values curiosity about AI/LLM patterns even without close prior experience.
Bring focused questions for the team
Write questions about how agents receive up-to-date code context, how generated changes are reviewed and evaluated, and how the team handles information security while collaborating on experiments.

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…