New opportunity

Auxo AI - Senior AI Engineer

AuxoAI · Bangalore, India

About this role

About the Role

AuxoAI is seeking a Senior AI Engineer to design and deploy production-grade AI agents capable of structured reasoning, planning, and decision-making. This role focuses on building intelligent agent systems and predictive ML solutions for real-world enterprise workflows. Responsibilities: - Design and architect modular AI agent frameworks, including skill decomposition, tool orchestration, and persistent state tracking. - Build and deploy supervised and unsupervised ML models for prediction, classification, anomaly detection, and pattern recognition. - Develop decision-making loops balancing exploration vs. exploitation, cost vs. accuracy, and latency vs. reasoning depth. - Create structured memory systems (episodic, semantic, and vector-based) with optimized retrieval. - Design tool-calling architectures with execution validation and failure recovery. - Develop evaluation frameworks using task success metrics, rollout simulations, and model accuracy benchmarks. - Integrate AI agents with enterprise systems (ERP APIs, data lakehouses). - Deliver production-ready systems meeting reliability, cost, and security standards. Requirements: - 3-8 years of experience in production machine learning or AI systems. - Hands-on experience with scikit-learn, XGBoost, or PyTorch. - Strong experience building or customizing agent frameworks. - Experience with cloud-native AI platforms (GCP Vertex AI, Gemini). - Proficiency in Python and scalable system design. - Experience with cloud-native infrastructure, APIs, containers, and distributed systems. - Strong understanding of RAG architectures and vector databases. - Experience with production-grade logging, monitoring, and debugging. Nice to Have: - Experience with Amazon Bedrock, SageMaker, Azure AI Foundry. - Familiarity with AI/LLM observability platforms (LangSmith, Langfuse, Grafana, Prometheus). - Experience with Kubernetes, Docker, Terraform, or Helm. - Knowledge of reinforcement learning, multi-agent systems, and MLOps. Note: Candidates who can join immediately or within 2 weeks are prioritized.

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

Skills for this role

Artificial IntelligenceMachine LearningPythonAgentic AIPyTorchGoogle Cloud PlatformGenerative AIMLOpsAWS Bedrockscikit-learnXGBoostGCP Vertex AIGeminiRAGVector DatabasesDatabricksDistributed SystemsKubernetesDockerTerraformHelmLangSmithLangfuseGrafanaPrometheus

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

Architecting Modular Agent Frameworks
The role requires designing systems that go beyond simple RAG, focusing on skill decomposition, tool orchestration, and persistent state tracking. You must demonstrate how to build frameworks that allow agents to plan and execute multi-step workflows reliably.
Balancing Operational Trade-offs
You are expected to manage complex decision-making loops that weigh exploration against exploitation, cost against accuracy, and latency against reasoning depth. Success depends on your ability to optimize these parameters for production-grade enterprise environments.
Building Robust Evaluation Frameworks
Because the role focuses on agentic systems, you must create rigorous evaluation pipelines. This includes task success metrics, rollout simulations, and multi-sample validation to ensure model accuracy and reliability before deployment.
Enterprise System Integration
You must integrate AI agents with complex enterprise infrastructure, including ERP APIs and data lakehouses. This requires strong Python engineering skills and experience with distributed systems to ensure secure, scalable, and maintainable production deployments.

Likely challenges to think through

Managing Agentic Failure Modes
Designing tool-calling architectures involves inherent risks of failure. You will need to prepare for discussions on how to implement execution validation, retry mechanisms, and graceful recovery strategies when agents fail to complete tasks.
Optimizing Memory Retrieval Strategies
Moving beyond basic vector search to structured memory (episodic and semantic) presents a technical challenge in retrieval optimization. Be prepared to discuss how you balance memory overhead with the need for low-latency, context-aware agent performance.
Production-Grade Observability
Maintaining visibility into agent reasoning and decision-making is difficult in distributed systems. You should be ready to explain how you use logging, monitoring, and tracing tools (like LangSmith or Prometheus) to debug non-deterministic agent behavior in production.

Practice your answers

Technical practice

Designing Modular AI Agent Frameworks

What to coverExplain how you would architect a system for skill decomposition and tool orchestration. Discuss your approach to maintaining persistent state across multi-step agent reasoning cycles and how you ensure modularity for future tool integration.

Balancing Exploration vs. Exploitation in Agent Loops

What to coverDescribe your strategy for managing the trade-off between exploration and exploitation in autonomous decision-making loops. Address how you balance reasoning depth against latency requirements and cost constraints in a production environment.

Structured Memory Systems for AI Agents

What to coverCompare the implementation of episodic, semantic, and vector-based memory stores. Explain how you would optimize retrieval strategies to ensure the agent accesses relevant context efficiently without exceeding token limits.

Tool-Calling Architecture and Failure Recovery

What to coverDetail your design for a robust tool-calling architecture. Focus on how you implement execution validation, retry mechanisms, and graceful failure recovery when an agent encounters an error during an external API call.

Evaluating Agent Performance

What to coverHow do you design an evaluation framework for AI agents? Discuss the use of task success metrics, rollout simulations, and multi-sample validation to ensure model accuracy and reliability before deployment.

Integrating AI with Enterprise Systems

What to coverDescribe the challenges of integrating AI agents with enterprise data sources like ERP APIs or data lakehouses. Focus on security, data consistency, and maintaining throughput when connecting agents to legacy infrastructure.

Production-Grade Observability for LLMs

What to coverExplain how you would implement logging, monitoring, and alerting for agentic systems. Suggest how tools like LangSmith, Langfuse, or Prometheus can be used to debug distributed agent workflows and track performance metrics.

Scalable System Design in Python

What to coverDiscuss your approach to writing scalable, maintainable Python code for AI systems. Address how you handle concurrency, containerization, and distributed system constraints when deploying models in production.

Behavioural practice

Managing Conflicting Project Requirements

What to coverDescribe a situation where you had to balance competing priorities, such as model accuracy versus cost efficiency. Explain the situation, the actions you took to evaluate the trade-offs, and the final result for the business.

Handling Production Failures

What to coverProvide an example of a time you had to troubleshoot a critical failure in a production ML system. Detail the steps you took to identify the root cause, your recovery strategy, and how you prevented the issue from recurring.

Collaborating on Complex Architectural Decisions

What to coverDescribe a time you proposed a specific architectural approach for an AI agent framework that was challenged by stakeholders. Explain how you communicated the technical benefits and reached a consensus to move forward.

Adapting to Rapidly Evolving AI Technologies

What to coverGive an example of how you have stayed current with the fast-paced changes in AI and LLM frameworks. Describe a specific instance where you applied a new technique or tool to improve an existing project's performance.

Sharpen your knowledge

Topics to prepare

AI Agent Architecture and Decision Loops
The role requires designing modular frameworks with skill decomposition and tool orchestration. Exercise: Draft a system design for an agent that manages a multi-step enterprise workflow, explicitly defining how it handles state tracking, tool-calling validation, and the trade-off between reasoning depth and latency.
Agent Evaluation Frameworks
The company emphasizes performance measurement beyond standard accuracy. Exercise: Design a testing suite for an AI agent that includes task success metrics, rollout simulations, and multi-sample validation to ensure reliability in production environments.
Structured Memory Systems for Agents
The role involves building episodic, semantic, and vector-based memory. Exercise: Compare and contrast retrieval strategies for these memory types and explain how you would implement a persistent state tracking mechanism for a long-running agent.

Skills to revise

Production-Grade ML Deployment
The role requires deploying models using scikit-learn, XGBoost, or PyTorch in enterprise settings. Exercise: Create a deployment pipeline script that includes automated logging, monitoring, and error handling for a model serving API.
Cloud-Native AI Integration
Experience with GCP Vertex AI and Gemini is preferred. Exercise: Outline how you would integrate a Gemini-based agent with an external ERP API or data lakehouse, focusing on security and scalable data retrieval.
Tool-Calling and Execution Validation
The role requires designing robust tool-use architectures. Exercise: Write a Python function that implements a tool-calling loop with a retry mechanism and a validation step to verify the output of an external API call before the agent proceeds.

Your preparation checklist

Prepare a technical walkthrough of an agentic system
Develop a concise presentation or document detailing a project where you built an agentic framework. Focus on how you handled skill decomposition, tool orchestration, and persistent state tracking, as these are core responsibilities for this role.
Demonstrate production-grade ML evaluation frameworks
Prepare examples of how you have measured model performance beyond simple accuracy. Be ready to discuss your experience with task success metrics, rollout simulations, and multi-sample validation in production environments.
Review cloud-native AI deployment patterns
Refresh your knowledge on deploying models via GCP Vertex AI and Gemini. Be prepared to discuss how you manage the trade-offs between latency, reasoning depth, and cost in a cloud-native, containerized architecture.
Formulate questions on enterprise system integration
Prepare thoughtful questions regarding their current data lakehouse or ERP integration challenges. Ask about how they handle execution validation and failure recovery when agents interact with these complex enterprise systems.
Document your experience with observability and MLOps
Compile specific examples of how you have implemented logging, monitoring, and alerting for AI systems. If you have used tools like LangSmith, Langfuse, or Prometheus, be ready to explain how these improved your debugging process.

Availability last checked: 18 Oct 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…