New opportunity

Senior AI Engineer

MaxIQ · Pune, India

About this role

MaxIQ seeks a Senior AI Engineer to architect autonomous agents for revenue operations in B2B SaaS. The agents will support sales, customer success and RevOps teams with predictive guidance, customer journey analysis and context-aware automation. The position lists Pune, India; Raleigh, United States; and San Francisco, United States as locations.

Responsibilities

Build agents using LangChain, LlamaIndex, and Python or Java to automate deal qualification and churn prediction, generate insights, and calculate customer health scores.
Design role-specific sales intelligence, including deal guidance for account executives and renewal alerts for customer success managers, with integrations such as Salesforce.
Advance chat-based assistants into autonomous systems using prompt engineering, retrieval-augmented generation and real-time data pipelines.
Implement tracing, logging and monitoring to make agent decisions observable in revenue-critical operations.
Optimize LLM inference for high-throughput use through quantization, model pruning and cloud-native deployment.

Qualifications

BS or MS in Computer Science, Data Science or a related field.
At least 5 years of experience with generative AI frameworks, LLM fine-tuning and agentic architectures; expertise in Python or Java, AWS or GCP, and containerized microservices.
Experience building observability into AI agents is a plus. The posting also seeks an interest in B2B SaaS revenue operations and a high-ownership startup environment.

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

Skills for this role

AI agentsLangChainLlamaIndexPythonJavaRetrieval-augmented generationPrompt engineeringLLM fine-tuningAWSGCPContainerized microservicesSalesforceReal-time data pipelinesTracingLoggingMonitoringQuantizationModel pruningCloud-native deploymentChurn predictionCustomer health scoring

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

Autonomous revenue agents
Architect agents for deal qualification, churn prediction, customer health scoring, and predictive guidance for sales, customer success, and RevOps teams.
Role-specific sales intelligence
Design deal guidance for account executives and renewal alerts for customer success managers, with integrations such as Salesforce.
RAG and real-time context
Advance chat-based assistants into autonomous systems using prompt engineering, retrieval-augmented generation, and real-time data pipelines.
Observable decisions
Implement tracing, logging, and monitoring so agent decisions are transparent in revenue-critical operations.
High-throughput inference
Optimize LLM performance through quantization, model pruning, and cloud-native deployment.

Likely challenges to think through

Autonomy versus control
A likely design challenge is deciding when an agent can act on revenue signals and when it should seek review. Prepare to explain how you would bound actions and make decisions inspectable; the posting does not specify an approval process.
Fresh data versus reliable guidance
Real-time customer data and RAG may create a trade-off between timely recommendations and incomplete or stale context. Prepare an approach to validating retrieved evidence and handling missing data; this is a preparation hypothesis, not a reported issue.
Throughput versus answer quality
Quantization and pruning can improve inference efficiency but may affect output quality. Prepare to describe how you would measure that trade-off for deal guidance or renewal alerts; the posting does not state performance targets.

Practice your answers

Technical practice

How would you architect an autonomous agent for a revenue-operations workflow?

What to coverSuggested preparation: Explain how you would define the agent’s goal, available data, actions and decision boundaries using LangChain or LlamaIndex with Python or Java. Discuss when autonomy adds value over a chat assistant and how you would evaluate its decisions.

How would you design an agent to support deal qualification?

What to coverSuggested preparation: Describe what inputs and outputs the workflow would need and how an agent could turn evidence into useful guidance for an account executive. Explain how you would test whether the guidance is accurate and actionable without treating an uncertain inference as fact.

How would you approach churn prediction and customer health scoring?

What to coverSuggested preparation: Distinguish a prediction of churn risk from a health score and explain how customer-journey information might inform each. Address data quality, evaluation and how to communicate uncertainty to customer success teams.

How would you deliver role-specific insights through an integration such as Salesforce?

What to coverSuggested preparation: Compare deal guidance for account executives with renewal alerts for customer success managers, including what context each needs. Explain how you would handle changing records and check that an insight is relevant to the intended user.

How would you use retrieval-augmented generation in an autonomous assistant?

What to coverSuggested preparation: Walk through retrieval, context selection, prompt construction and validation of a generated response or proposed action. Discuss trade-offs between relevant current context, latency and the risk of unsupported conclusions.

How would real-time data pipelines affect agent decisions?

What to coverSuggested preparation: Explain how you would keep customer-journey context current enough for predictive guidance and context-aware automation. Discuss how you would detect stale or missing inputs and what the agent should do when reliable context is unavailable.

When would you use prompt engineering rather than LLM fine-tuning?

What to coverSuggested preparation: Compare the two approaches for improving an agent’s revenue-operations outputs, including evaluation effort and maintenance as business context changes. Explain what evidence would justify moving beyond a prompt-based approach.

What would you trace and monitor to make agent decisions observable?

What to coverSuggested preparation: Describe how tracing, logging and monitoring could connect an agent’s inputs, retrieval, decisions and outputs. Explain how you would use that evidence to investigate a questionable recommendation in a revenue-critical workflow.

How would you improve LLM inference throughput without losing useful output quality?

What to coverSuggested preparation: Compare quantization and model pruning as options named in the posting, including their potential effects on latency, resource use and output quality. Describe how you would benchmark the trade-off against the agent’s tasks.

How would you deploy and scale these agents as containerized microservices?

What to coverSuggested preparation: Outline how you would separate agent services and inference workloads for cloud-native deployment on AWS or GCP. Discuss scaling for high-throughput use and how deployment choices interact with observability and reliability.

Behavioural practice

Tell me about a time you took ownership of an ambiguous technical problem.

What to coverSuggested preparation: Use a situation–action–result example showing how you defined the problem, made decisions and adapted as you learned more. Relate the example to the posting’s emphasis on high ownership in a fast-paced environment.

Tell me about a time you aligned technical work with the needs of different users.

What to coverSuggested preparation: Use a situation–action–result example that shows how you learned what distinct users needed and resolved competing priorities. The role’s sales, customer success and RevOps audiences make role-specific communication particularly relevant.

Describe a time observability helped you investigate an unreliable system.

What to coverSuggested preparation: Use a situation–action–result example covering the signals you examined, the action you took and what changed. Focus on how you made a consequential system’s behaviour understandable rather than assuming any particular incident at MaxIQ.

Tell me about a time you made a quality-versus-speed trade-off.

What to coverSuggested preparation: Use a situation–action–result example explaining the constraints, how you evaluated options and how you checked the outcome. This is relevant to the posting’s combination of high-throughput inference and revenue-critical agent decisions.

Sharpen your knowledge

Topics to prepare

Autonomous agent architecture
Sketch how an agent moves from a customer event to a decision, tool use and a final recommendation; mark where it should stop or seek review. Relevant because the role architects autonomous systems for revenue operations.
B2B SaaS revenue signals
Define, on paper, the inputs and limitations of deal qualification, churn prediction and customer health scores, then explain how an account executive’s guidance would differ from a customer success manager’s renewal alert. These are stated agent use cases.
Prompt engineering, retrieval and fresh data
Diagram how a prompt, retrieved customer context and a real-time update would produce an answer; identify what happens when sources disagree or are stale. The posting calls for prompt engineering, RAG and real-time data pipelines.
LLM fine-tuning and inference trade-offs
Compare when fine-tuning, retrieval, quantization or model pruning would address a given quality, latency or throughput problem. Fine-tuning is listed in the qualifications; quantization and pruning are named inference responsibilities.
Agent observability
Trace a hypothetical incorrect renewal alert backward through its inputs, retrieval and agent decisions, and specify what logs and monitoring signals would help diagnose it. The role calls for transparent decisions in revenue-critical operations; prior agent-observability experience is described as a plus.
Cloud-native service design
Draw the boundaries between an agent service, its data pipeline and its model inference service, then reason about scaling and failure handling. The posting specifies AWS or GCP expertise, containerized microservices and cloud-native deployment.

Skills to revise

Build a small revenue agent
As a preparation exercise, use Python or Java with LangChain or LlamaIndex to process fictional account records and produce a deal-qualification recommendation with its supporting inputs. These languages and frameworks are explicitly named; the exercise is a suggestion, not an employer assignment.
Prototype role-specific CRM output
Using fictional Salesforce-style records, draft a deal-guidance response for an account executive and a renewal alert for a customer success manager; check that each uses the appropriate fields. The posting describes these outputs and integration with tools like Salesforce.
Implement a retrieval-backed workflow
As a practice exercise, retrieve from a small fictional customer-history set, incorporate a newly arrived event and inspect whether the resulting answer reflects current evidence. This rehearses the posting’s RAG and real-time pipeline work.
Instrument agent decisions
Add traces and structured logs to a prototype agent, then define monitoring signals for failed retrievals, missing inputs and decision latency. Tracing, logging and monitoring are explicit responsibilities.
Test inference optimizations
On a sample LLM workload, compare baseline and quantized inference for throughput, latency and output quality; outline how you would evaluate pruning before adopting it. The role explicitly calls for high-throughput inference optimization using quantization and model pruning.
Package a cloud-ready service
Containerize a small agent API and explain how you would deploy and scale it on AWS or GCP without assuming a specific service. Containerized microservices, AWS or GCP, and cloud-native deployment are stated in the posting.

Your preparation checklist

Prepare an agent project walkthrough
The role calls for autonomous agents that support deal qualification, churn prediction, and customer health scoring. Prepare a relevant project walkthrough, if available, covering the agent’s decisions, data inputs, outcomes, and your contribution; otherwise, label a proposed example as a design exercise.
Sketch a revenue-operations agent
The posting names sales, customer success, and RevOps users, with role-specific guidance and integrations such as Salesforce. Draw a proposed workflow showing how an agent would produce deal guidance for an account executive or a renewal alert for a customer success manager.
Design the retrieval and data flow
The role calls for prompt engineering, retrieval-augmented generation, and real-time data pipelines. Prepare a diagram showing data ingestion, retrieval, prompts, agent actions, and how you would check the relevance of generated guidance.
Show how agent decisions would be observable
The posting specifies tracing, logging, and monitoring in revenue-critical operations. Prepare a sample decision trace and identify the signals you would monitor to investigate incorrect or unexpected guidance.
Explain inference trade-offs
The role names high-throughput inference, quantization, model pruning, and cloud-native deployment. Prepare a short comparison of when you would use each approach and how you would measure latency, throughput, and output quality.
Gather evidence for the technical requirements
The posting specifies 5+ years with generative AI frameworks, LLM fine-tuning, and agentic architectures, plus Python or Java, AWS or GCP, and containerized microservices. Organize concrete examples you can accurately discuss for each relevant area; identify any gaps rather than implying experience you do not have.
Prepare questions about scope and success
Ask which revenue-operations workflows are the initial priority, what data and Salesforce integration are available, and how the team evaluates agent guidance, observability, and inference performance.

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…