GenAI Architecture

- Design and deliver Generative AI and agentic architectures on AWS using Amazon Bedrock and SageMaker AI, including prompt engineering, guardrails, tool/function calling, memory, evaluation, and multiagent patterns.
- Define model selection strategies (foundation vs finetuned vs hosted OSS models) with latency, cost, and quality tradeoffs.
- Establish GenAI reference architectures for enterprise adoption.

RAG & Knowledge Systems

- Build RetrievalAugmented Generation (RAG) systems using:
- Vector stores: Amazon OpenSearch (vector engine), Aurora PostgreSQL (pgvector), DynamoDB (vector search), Pinecone
- Embedding strategies, chunking, metadata enrichment, reranking, caching, and freshness controls
- Implement hallucination mitigation techniques, citation grounding, and response evaluation.

Modernization & CloudNative Architecture

- Modernize legacy workloads into microservices, eventdriven, and serverless architectures using:
- EKS / ECS / Lambda
- API Gateway, EventBridge, SQS, SNS
- Apply domaindriven design (DDD) and integration best practices.
- Design and expose APIs using REST and GraphQL patterns (including schema design, resolver performance, and backend integration) based on usecase requirements.

MLOps / LLMOps

- Establish ML and GenAI CI/CD pipelines, including:
- Versioned prompts, models, datasets




- Evaluation pipelines (quality, safety, cost)
- Monitoring, drift detection, approval, rollback
- Implement feature engineering and feature management using SageMaker AI (integrated feature store capabilities).
- Support A/B testing and canary deployment of GenAI solutions.

Security, Governance & Responsible AI

- Define endtoend security architecture for GenAI:
- IAM, KMS, secrets management
- Data privacy, PII handling, prompt logging
- Enforce Responsible AI controls: guardrails, moderation, auditability, traceability.
- Align solutions with enterprise governance and compliance standards.

Reliability, Cost & Performance

- Define NFRs, SLIs/SLOs for GenAI (latency, throughput, cost per request).
- Implement observability using CloudWatch, XRay, OpenTelemetry.
- Optimize inference cost using caching, batching, routing, and model selection policies.

Stakeholder Engagement & Enablement

- Lead architecture reviews and design forums.
- Mentor engineers and architects across teams.
- Support solutioning, estimation, PoCs, and RFP responses for GenAI initiatives.

Core Skills & Qualifications (Refined)

AWS Core

Musthave

- VPC, IAM, KMS, CloudWatch, CloudTrail
- S3, DynamoDB, RDS/Aurora
- EKS / ECS / Lambda
- API Gateway, EventBridge, SQS, SNS

📌 AI Architect (India)
🏢 Tata Consultancy Services
📍 India