AI Backend Engineer (SDE-2)
Company: plotlineso
Location: Bengaluru Office
Type: Full-time
Posted: 2026-06-27
About this role
## Overview
Join our rapidly growing AI startup as an AI Backend Engineer and build the agentic core of our platform: the LLM workflows, agent orchestration, and AI-driven systems that turn natural-language marketer intent into live, personalized in-app experiences at scale. This role is for an engineer who pairs strong backend fundamentals with hands-on applied AI, and who wants to own systems end to end on a high-throughput production hot path.
Plotline is backed by Elevation Capital and supports customers like Dream11, Upstox, and Zepto.
## Why This Role Matters
Build the AI layer that turns marketer intent into live in-app experiences for 300M+ end users.
Work directly with the engineering team and founders to shape the agentic core of the product.
Design backend systems and patterns that scale with our high-growth trajectory.
We're seeking a rare combination of backend rigor, applied AI judgment, hands-on execution, and the drive to own systems end to end.
## About This Role
Key challenges you'll solve:
Agent Orchestration: Building LLM agents that plan multi-step and use tools reliably, deciding which steps to take rather than following a hardcoded script.
Production Reliability: Keeping a non-deterministic AI system fast, safe to retry, and observable on a sub-100ms hot path serving millions of devices.
This is an opportunity to have outsized impact on the core of the product while building deep expertise in agentic systems at a scaling startup.
## Outcomes Expected
Own and scale both our core backend and the agentic core, with the data modeling, resilience, and cost controls needed to run them reliably in production at scale.
Ship grounded, tool-calling agent flows end to end, backed by retrieval and a real evaluation harness.
## Responsibilities
Agent & AI Systems
LLM Orchestration: Design agents with tool-calling and real multi-step planning that select tools and recover when a tool fails or returns nothing useful.
Retrieval (RAG): B...