Personal details
Krishna V.
About
• Experienced AI professional and researcher with extensive experience in implementing projects across Generative AI, Machine Learning, Agentic AI, Deep Learning, Computer Vision, Natural Language Processing, Predictive Analytics, and Conversational AI.
• Currently working with Microsoft on designing and implementing solutions using LLM, AI Agents, MCP and driving ec2 advancements in NLP, Gen AI for the organization.
• Designed and deployed scalable AI solutions using AWS Bedrock, Azure Foundry utilizing its managed service for deploying and running generative AI models, enabling faster and more efficient model deployment.
• Developed and deployed scalable models leveraging LoRA and QLoRA for natural language processing tasks, improving accuracy and efficiency in real-world applications.
• Managed and optimized large datasets within Google BigQuery, leveraging its distributed architecture to ensure fast and cost-effective data analysis for AI-driven applications.
• Implemented Guardrails for AI models to ensure compliance with privacy regulations, including automatic PII redaction, to protect sensitive information during model inference.
• Utilized LangGraph to visualize agent behavior and logic flow, ensuring transparency and effective management of decision trees for complex LLM-based agents.
• Developed and deployed Generative AI models using TensorFlow, PyTorch, and Keras, creating high-quality text, image, and audio generation systems for various applications.
• Built and maintained Python-based AI services using LangChain and CrewAI, implementing RAG-based retrieval and Agentic AI workflows to enhance data retrieval and decision-making processes.
• Integrated OpenAI, Bard, Claude, and Azure OpenAI APIs, optimizing model performance by fine-tuning temperature, top-p, and max tokens, reducing hallucinations through embedding-based retrieval
Work Experience
Orchestrated multi-agent workflows using Microsoft Agent Framework, building production-grade AI agents on Azure Foundry with Azure Functions for stateful coordination and Cosmos DB vector search. Integrated Azure SRE Agent (Preview) for autonomous platform monitoring, configuring KQL queries, dynamic thresholds, and incident automation across AKS and App Service with human-in-the-loop approvals.
Engineered MCP-standardized agent-to-agent (A2A) communication, enabling autonomous handoffs between Claude and Azure OpenAI agents (triage to analysis to resolution). Developed modular backend services on Azure Container Apps, exposing secure APIs via Azure API Management for enterprise agentic pipelines.
Implemented observability with Azure Monitor and Log Analytics, defining SLOs/SLIs and leveraging SRE Agent for explainable root cause analysis, reducing toil and improving MTTR. Built RAG pipelines using Cosmos DB vector search, enabling agents to retrieve historical incidents and configurations, reducing hallucinations by 95% in triage workflows.
Designed privacy guardrails with Azure Content Safety and Purview, including PII redaction pipelines to ensure compliance in agent-based inference. Automated CI/CD and infrastructure with Azure DevOps (Bicep) and GitHub Actions, enabling zero-downtime deployments with rollback strategies tied to system health metrics.
Optimized real-time LLM inference via Azure OpenAI endpoints, using intelligent routing for cost and performance-aware multi-model orchestration. Established SRE best practices including error budgets, proactive alerting, and automated health summaries. Collaborated with platform and security teams to standardize agent integration patterns, accelerating internal adoption by 3x.
Developed Agentic and Generative AI solutions at Toyota North America using Gemini, GPT-4, and Claude to build scalable applications addressing real-world business needs and meeting performance SLAs. Designed modular backend services and RESTful APIs with Python (FastAPI), emphasizing reliability, extensibility, and best practices such as code reviews and version control.
Orchestrated autonomous agent workflows using LangChain and LangGraph, integrating prompt chaining, RAG, and LLM calls to deliver enterprise-grade AI features and simplify adoption for non-AI teams. Containerized and deployed microservices using Docker and CI/CD pipelines, optimizing resource usage across AWS Lambda and SageMaker for scalable applications.
Built secure data pipelines and endpoints with encryption, IAM policies, and PII redaction to meet enterprise security and compliance standards. Integrated vector databases (Pinecone, Qdrant) and semantic retrieval to improve RAG efficiency and context-aware performance.
Deployed serverless solutions using AWS Lambda and Bedrock for real-time inference and agent orchestration, balancing latency and cost. Established observability with AWS CloudWatch, tracking metrics, logs, and traces to ensure SLA adherence and system reliability.
Developed metrics pipelines to evaluate model and retrieval performance, enabling data-driven improvements and measurable impact on KPIs such as efficiency, adoption, and retention. Applied LangGraph for graph-based data modeling, supporting knowledge graphs and enhancing agent reasoning.
Automated deployment, versioning, and rollback using Infrastructure as Code (AWS CloudFormation), standardizing delivery workflows. Collaborated cross-functionally with engineering, product, and security teams in agile environments, translating business needs into scalable solutions and mentoring junior engineers.