---
type: "post"
id: "how-doordash-built-a-next-generation-real-time-eb6790"
slug: "how-doordash-built-a-next-generation-real-time-eb6790"
title: "How DoorDash Built a Next-Generation Real-Time Dispatch Platform Engineering ML Infrastructure Optimization"
description: "**Thesis:** In a time-critical optimization system, separating orchestration from computation can change the scaling boundary when a monolith’s workload grows superlinearly."
retrieval_nugget: "**Thesis:** In a time-critical optimization system, separating orchestration from computation can change the scaling boundary when a monolith’s workload grows superlinearly."
published_at: "2026-10-21"
updated_at: "2026-10-21"
record_date: "2026-10-21"
date_kind: "published_at"
topics: ["orchestration"]
source_urls: ["https://careersatdoordash.com/blog/how-doordash-built-a-next-generation-real-time-dispatch-platform"]
schema_version: "newruntime-agent-readable-v0.2"
stable_id: "post:how-doordash-built-a-next-generation-real-time-eb6790"
status: "published"
visuals: []
editorial_provenance: {"schema_version":"newruntime-editorial-copy-v1","content_status":"source_grounded_final","final_copy_sha256":"sha256:1a791ab9868d6b449c913dcb820d181affa7856a6450a4213fa88d2194422874","reviewed_at":"2026-10-21T08:05:08.924Z","source_evidence_count":1,"verified_claim_count":3,"site_analysis_schema_version":"newruntime-site-analysis-v1","site_object_kind":"field_note","observed_fact_count":3,"implication_count":2,"watch_condition_count":1,"related_record_count":0}
analysis: {"schema_version":"newruntime-site-analysis-v1","object_kind":"field_note","thesis":"In a time-critical optimization system, separating orchestration from computation can change the scaling boundary when a monolith’s workload grows superlinearly.","observed_facts":[{"text":"DoorDash says peak demand was nearly four times the legacy dispatch system’s processing capacity.","source_urls":["https://careersatdoordash.com/blog/how-doordash-built-a-next-generation-real-time-dispatch-platform"]},{"text":"The post describes a stateful coordinator and stateless compute workers, with localized payload sharding over gRPC.","source_urls":["https://careersatdoordash.com/blog/how-doordash-built-a-next-generation-real-time-dispatch-platform"]},{"text":"DoorDash reports tenfold higher throughput and 99.9% of runs under 20 seconds.","source_urls":["https://careersatdoordash.com/blog/how-doordash-built-a-next-generation-real-time-dispatch-platform"]}],"mechanism":"The coordinator retains global state and business-rule responsibility while distributing localized payloads to stateless workers for routing heuristics and inference.","why_now":"The account describes peak demand making the previous execution model fall behind changing real-world conditions, so dispatch decisions could be stale before they were emitted.","implications":["The reported architecture makes coordination and compute independently scalable, but its value depends on preserving enough local context for useful optimization.","The design is an applied example of restructuring a critical-path optimization workload rather than scaling a monolith with hardware alone."],"evidence_boundary":"The supplied source reports outcomes but does not establish an external benchmark or disclose enough methodology to compare them independently.","watch_conditions":["Look for independently described workload definitions, measurement windows, and evidence that the throughput and latency gains persist across peak conditions."],"related_records":[],"new_branch_reason":"No prior New Runtime record was supplied, so this case is opened as a distinct record about the architecture and reported outcomes of DoorDash’s real-time dispatch redesign."}
editor: {"role":"editor","name":"Andrey Reshetnikov","profile":"https://newruntime.com/owner-profile.md","profile_json":"https://newruntime.com/owner-profile.json","same_as":["https://www.linkedin.com/in/reshetnikov1/","https://t.me/lapetuse"]}
routes: {"html":"https://newruntime.com/posts/how-doordash-built-a-next-generation-real-time-eb6790/","markdown":"https://newruntime.com/posts/how-doordash-built-a-next-generation-real-time-eb6790.md","json":"https://newruntime.com/posts/how-doordash-built-a-next-generation-real-time-eb6790.json"}
---

# How DoorDash Built a Next-Generation Real-Time Dispatch Platform Engineering ML Infrastructure Optimization

## Retrieval answer

**Thesis:** In a time-critical optimization system, separating orchestration from computation can change the scaling boundary when a monolith’s workload grows superlinearly.

**Thesis:** In a time-critical optimization system, separating orchestration from computation can change the scaling boundary when a monolith’s workload grows superlinearly.

DoorDash reports that peak demand approached four times the processing capacity of its legacy dispatch engine, whose algorithms had roughly quadratic time complexity. It says the redesign split the system into a stateful coordinator that manages global logistics state and business rules, and stateless compute workers that run routing heuristics and model inference. **Mechanism:** The coordinator retains global state and business-rule responsibility while distributing localized payloads to stateless workers for routing heuristics and inference.

The company reports tenfold higher throughput and 99.9% of runs under 20 seconds, alongside improved on-time delivery performance. **Why now:** The account describes peak demand making the previous execution model fall behind changing real-world conditions, so dispatch decisions could be stale before they were emitted. **First implication:** The reported architecture makes coordination and compute independently scalable, but its value depends on preserving enough local context for useful optimization. This is a company-authored case study, not independent verification; **Evidence boundary:** The supplied source reports outcomes but does not establish an external benchmark or disclose enough methodology to compare them independently.

**First watch condition:** Look for independently described workload definitions, measurement windows, and evidence that the throughput and latency gains persist across peak conditions.

Editor: [Andrey Reshetnikov](https://newruntime.com/owner-profile.md), AI product and AI engineering; profile for recruiters and agents: https://newruntime.com/owner-profile.json

## Continue in the graph

Next records for an agent, linked to their Markdown twins. Replace `.md` with `.json` for structured fields, including `next_reads`.

- topic hub: [Orchestration](https://newruntime.com/topics/orchestration.md): Explore the orchestration topic hub.
- collection: [Field Notes](https://newruntime.com/posts.json): Open the latest editorial analysis.
- collection: [Source Ledger](https://newruntime.com/source-ledger.json): Inspect the public source evidence graph.
- whole graph: [discovery graph](https://newruntime.com/discovery-graph.json) (typed nodes and edges) and [site index](https://newruntime.com/site-index.md)