The Fifth Elephant
The Fifth Elephant

The Fifth Elephant Pune edition

Call for Proposals

All submissions

Sathish

@sathish316

Pied-Piper: Orchestrating Coding Agents for Long-running/Repeatable SDLC workflows

Submitted Jan 9, 2026

Indicate the track in which your submission fits: Track 1 AI in Software Development Life Cycle (SDLC) Type of submission: 30 mins talk

Session Description

Coding Agents started as IDE autocomplete at the start of 2025 and have become more mainstream in the programming community by the end of 2025 through various innovations like Cursor Tab model, Claude Code CLI etc. The big shift came when some programmers figured out ways to run multiple Coding CLIs and orchestrate them to build complex SDLC workflows using various approaches.

In this talk, we will discuss a Coding Agents Orchestration framework called Pied-Piper. Pied-Piper orchestrates Coding CLIs like Claude Code, Rovo Dev using the concept of Teams, Subagents, Playbooks. It creates a team of Subagents with defined roles and responsibilities and a declarative workflow. Subagents behave like Team-mates working together using a shared Todo list and Agentic memory layer in the form of Beads Task manager. Subagents follow a specific plan that is similar to Jira workflows and can do this repeatedly.

It helps us in leveraging a team of Coding Agents for long-running and repetitive SDLC workflows like:

  1. Test coverage improvement
  2. Language migrations
  3. Techstack migrations
  4. Microservice consolidation to reduce Service Sprawl etc

These workflows are modeled as Playbooks using Subagent roles with defined responsibilities and some of the successful playbooks we have customized are explained in the open-source repo. Using a Task memory layer and Subagents based orchestration helps in running Coding agents for long hours without running out of context window limits. This is because of the way Subagents are designed to not share the parent context for every invocation.

Pied-Piper is built with the understanding that orchestrating multiple coding agents needs to have Human-in-the-loop approvals/reviews to maintain code quality and reliability. We are internally leveraging Pied-Piper for large-scale refactoring of multiple microservices and consolidating it into fewer services. It is open-sourced under MIT License - https://github.com/sathish316/pied-piper

Attached Mindmap for outline of the Talk - https://drive.google.com/file/d/1fxlz7kNuMh3ZzBZ8pG_lep5mD_lr6AXw/view?usp=sharing

Key Takeaways

  • Orchestrating multiple coding agents
  • Playbooks for leveraging coding agents for repeatable work
  • Enforcing quality guardrails with maker-checker model and human-in-the-loop approvals
  • Lessons from orchestrating Coding Agents for real-world usecases

Target Audience

  • Software Engineers
  • AI/ML Engineers

Bio

Sathish Kumar is a Senior Principal Engineer at Atlassian working on problems in ITOps, Alerts/Incidents management products, AIOps. He has spent close to 20 years in the industry working on large-scale distributed systems, ML ranking systems and b2c/b2b products/platforms.

All submissions

Comments

{{ gettext('Login to leave a comment') }}

{{ gettext('Post a comment…') }}
{{ gettext('New comment') }}
{{ formTitle }}

{{ errorMsg }}

{{ gettext('No comments posted yet') }}

Hosted by

The Fifth Elephant
The Fifth Elephant
Jumpstart better data engineering and AI futures