About Me

Staff Engineer | Software Architect

Yogesh Badke

What I believe about AI in production systems

Most “AI for DevOps” demos fail for the same reason: they hand an LLM a shell and hope. That's not autonomy : it's a liability with a chat interface.

Agents become trustworthy when you give them three things the demos skip: declarative abstractions they can reason over, a dependency graph they can walk, and deterministic execution for anything that touches production. The LLM proposes; the platform disposes. Humans approve the shape, not every keystroke.

This is the pattern behind the arc I built at Dream11: Odin as the substrate (shipped, open source, running IPL traffic), Sentinel as the agent that reads from it (~60% RCA accuracy : a floor I expect to embarrass in eighteen months), and Asgard as the agent that writes to it (in progress). The bottleneck is rarely the model. It's the scaffolding around it.

The older work : Optimus, Scaler, the 14-minute CI/CD pipeline, the 70% bug reduction : taught me the scaffolding. The current work is about handing the keys to agents without handing them the company.

Over 14 years building high-scale platforms : distributed systems, microservices architecture, and engineering efficiency.

What I Do

  • Microservices & Distributed Systems
  • Multi-tenant Platform Architecture
  • Auto-scaling & Performance Optimization
  • DevOps & CI/CD Pipeline Design
  • Database-as-a-Service Platforms

Technical Stack

Platform & Backend

Java Python Go Spring Boot Vert.x

Infrastructure

Kubernetes Terraform AWS Docker

Data

PostgreSQL MySQL Redis Kafka Cassandra

AI & Agents

LangGraph ReAct Tree-of-Thought

Observability

Datadog Grafana Elasticsearch