New opportunity

AI Lead Architect

Dentsu Global Services Private Limited · Pune, India

About this role

The AI Lead Architect will spearhead hands-on engineering for end-to-end AI solutions, focusing on Deep Learning, Generative AI, Agentic AI, and multimodal use cases. The role involves defining reference architectures, conducting solution design reviews, and ensuring delivery excellence across client engagements. Responsibilities include designing multi-agent systems with reasoning, planning, and persistent memory, as well as establishing patterns for SLM design, fine-tuning, and deployment. The architect will also define evaluation frameworks, enforce safety and guardrail standards, and oversee production readiness, including monitoring and drift detection. Key duties involve mentoring engineers, enforcing AI in SDLC frameworks, and supporting pre-sales activities. Required qualifications include a minimum of 8 years of total software development experience, with at least 3 years specifically in building and deploying Deep Learning/AI systems in production. Candidates must demonstrate hands-on experience with multi-agent systems, custom fine-tuning, and advanced RAG. Proficiency in Python, SQL, and backend frameworks like FastAPI or Django is mandatory, along with experience in cloud platforms (AWS, Azure, or GCP) and various database technologies. Preferred qualifications include experience with commerce cloud ecosystems such as Salesforce and Adobe. The role requires working from the Pune office at least 3 days per week.

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

Skills for this role

Deep LearningMachine LearningNeural NetworksTransformersPredictive ModelingEmbeddingsVector SearchGenerative AILLMsSLMsRAGAgentic RAGAI AgentsPrompt EngineeringMultimodal ArchitecturesSFTLoRAQLoRARLHFRLAIFDistillationQuantizationLangGraphLlamaIndexAutoGenPythonSQLFastAPIFlaskDjango

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 Multi-Agent Systems
The role requires designing complex, agentic workflows that incorporate reasoning, planning, tool use, and persistent memory. You must demonstrate the ability to build systems that go beyond simple RAG, focusing on inter-agent communication and grounded retrieval using frameworks like LangGraph or AutoGen.
Production-Grade AI Lifecycle Management
You are expected to establish evaluation frameworks, guardrails, and observability for AI systems. This includes managing model drift, latency, and cost, as well as enforcing safety standards through red-teaming and adversarial testing in enterprise environments.
Strategic Model Selection and Optimization
A core responsibility is performing trade-off analysis between frontier models and SLMs. You must be prepared to implement patterns for distillation, fine-tuning, and quantization to meet specific enterprise constraints regarding data residency, cost, and deployment environments.
Technical Leadership and Mentorship
Beyond individual contribution, you must act as a technical sentinel, mentoring engineers and enforcing AI-specific SDLC frameworks. This involves conducting rigorous architecture reviews and communicating complex technical risks to non-technical stakeholders.

Likely challenges to think through

Balancing Innovation with 'Fail Fast' Rigor
The role demands a high-speed, iterative approach while simultaneously requiring the discipline to disqualify unviable use cases early. Prepare to discuss how you objectively evaluate technical feasibility and risk to prevent resource wastage on projects that do not meet enterprise standards.
Managing Diverse Stakeholder Expectations
You will bridge the gap between complex AI architecture and client business needs. Prepare to explain how you translate technical constraints—such as model latency or hallucination risks—into clear, actionable insights for non-technical leadership during pre-sales and delivery phases.
Ensuring Reliability in Regulated Environments
Deploying agentic systems in enterprise settings introduces significant safety and compliance hurdles. Prepare to discuss your experience with adversarial testing and building automated evaluation suites that ensure consistent performance and hallucination control in high-stakes production scenarios.

Practice your answers

Technical practice

Architecting Multi-Agent Systems

What to coverDiscuss your experience designing multi-agent systems using frameworks like LangGraph or AutoGen. Focus on how you handle inter-agent communication, persistent memory, and planning logic to ensure reliable task execution.

SLM Design and Optimization Trade-offs

What to coverExplain your approach to choosing between frontier models and Small Language Models (SLMs). Address how you implement distillation, quantization, and fine-tuning techniques to meet specific enterprise constraints regarding latency, cost, and data residency.

Advanced RAG and Knowledge Architectures

What to coverDescribe how you design hybrid retrieval systems that integrate vector, graph, and NoSQL databases. Explain the role of Knowledge Graphs in improving grounding and entity linking for complex generative AI applications.

Production AI Evaluation Frameworks

What to coverDetail how you establish evaluation as a first-class discipline in production environments. Discuss your methodology for creating golden datasets, automated regression suites, and human-in-the-loop feedback loops to monitor model drift and performance.

Safety, Guardrails, and Adversarial Testing

What to coverExplain your strategy for enforcing safety and hallucination-control standards. Describe how you conduct red-teaming and adversarial testing to ensure high-stakes deployments remain secure and reliable.

Fine-tuning Techniques and Lifecycle Management

What to coverCompare SFT, LoRA, and QLoRA in the context of enterprise deployment. Discuss how you manage the lifecycle of fine-tuned models, including versioning, serving, and GPU resource optimization.

Multimodal System Design

What to coverDiscuss the challenges of building systems that process text, vision, and speech. Focus on your experience with data ingestion, representation, and ensuring downstream agent reasoning remains consistent across different modalities.

AI in SDLC and Production Readiness

What to coverHow do you integrate AI development into standard SDLC frameworks? Explain your criteria for production readiness, specifically regarding monitoring, incident response, and infrastructure automation across cloud environments.

Behavioural practice

Disqualifying Unviable AI Use Cases

What to coverDescribe a situation where you had to apply 'fail fast' logic to an AI project. Explain how you evaluated technical feasibility, risk, and cost to justify discontinuing the project early in the cycle.

Communicating Technical Risk to Stakeholders

What to coverProvide an example of how you explained complex AI architectural risks or feasibility issues to non-technical stakeholders. How did you ensure they understood the trade-offs involved in your decision-making?

Mentoring and Capability Building

What to coverDescribe a time you mentored engineers through a difficult architecture or code review. How did you ensure the process resulted in permanent capability building rather than just solving an immediate crisis?

Managing Conflicting Client Requirements

What to coverShare an experience where you had to balance competing client needs during a solution design review. How did you facilitate a subjective technical decision while maintaining delivery excellence?

Sharpen your knowledge

Topics to prepare

Multi-Agent System Architecture
The role requires designing systems with reasoning, planning, persistent memory, and inter-agent communication. Exercise: Draft a high-level architecture diagram for a multi-agent system where one agent handles retrieval, another performs reasoning/planning, and a third manages tool execution, ensuring you define the state management and memory persistence strategy.
SLM Design and Optimization
Focus on enterprise constraints like cost, latency, and data residency. Exercise: Create a decision matrix comparing the trade-offs between using frontier models versus fine-tuned Small Language Models (SLMs) for a hypothetical high-volume enterprise application, considering distillation, quantization, and routing strategies.
AI Evaluation and Guardrails
Establishing evaluation as a first-class discipline is a core responsibility. Exercise: Design a comprehensive evaluation framework for an agentic RAG system that includes golden datasets, automated regression suites, and a plan for human-in-the-loop validation to control hallucinations.

Skills to revise

Agentic Framework Implementation
Hands-on experience with LangGraph, LlamaIndex, or AutoGen is required. Exercise: Build a simple workflow using LangGraph that demonstrates a multi-step agentic process, such as a loop that performs a search, evaluates the results, and decides whether to refine the query or provide a final answer.
Advanced RAG and Knowledge Architectures
The role involves hybrid retrieval spanning vector, graph, and NoSQL stores. Exercise: Implement a prototype that combines vector search with graph-based retrieval (KG-assisted) to improve entity linking and grounding for a complex domain-specific query.
Fine-tuning and Model Optimization
Practical experience with SFT, LoRA/QLoRA, and quantization is mandatory. Exercise: Prepare a script using a library like PEFT to perform LoRA fine-tuning on a small open-source model, documenting the memory savings achieved through quantization and the impact on inference latency.

Your preparation checklist

Prepare a Multi-Agent System Case Study
Develop a technical walkthrough of a production-grade multi-agent system you have built. Focus on how you implemented reasoning, planning, persistent memory, and tool use. Be prepared to discuss the specific orchestration framework used (e.g., LangGraph, AutoGen) and how you handled inter-agent communication.
Document Your SLM and Fine-Tuning Experience
Create a summary of your experience with Small Language Models (SLMs). Highlight specific projects where you applied distillation, quantization, or fine-tuning (SFT, LoRA/QLoRA) to meet constraints like latency, cost, or data residency. Explain your decision-making process for choosing between frontier models and SLMs.
Design an AI Evaluation and Guardrail Framework
Outline a comprehensive evaluation strategy for generative AI systems. Include how you define golden datasets, implement automated vs. human-in-the-loop testing, and establish guardrails to prevent hallucinations. Prepare to explain how you monitor for drift and ensure production reliability in regulated environments.
Formulate Your 'Fail Fast' Methodology
Prepare to discuss your approach to evaluating AI use cases for technical feasibility, risk, and ROI. Be ready to provide an example of a project you recommended against pursuing and the specific trade-off analysis (cost, effort, risk) that led to that decision.
Prepare Strategic Questions for the Interview
Draft thoughtful questions regarding the team's current AI SDLC frameworks, the types of commerce cloud integrations (Salesforce/Adobe) they prioritize, and how they balance pre-sales architectural framing with long-term delivery excellence.

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…