---
title: "DE Dost: Teaching Machines to Be a Friend on the Last Mile"
description: "DE Dost is a GenAI onboarding companion embedded in Swiggy’s delivery-executive app, designed for new DEs who receive information but lack confidence-building, immediate guidance. It supports questions in nine languages across SOPs, earnings, payouts, app navigation, shifts, and support, adapting detail and tone to conversational context. Before the first LLM call, a graph state assembles user context, preprocessing resolves references, and embedding-based semantic routing separates general knowledge-base questions from personal diagnostics. Retrieval uses authored QnA pairs plus hybrid semantic and lexical scoring to handle paraphrases and terms such as “MG” and “COD,” while guardrails constrain follow-ups to available knowledge. Evaluation spans about 800 cases per language across fleet and payout combinations, with macro accuracy trending at 90%+ and a target P50 latency of 4.5 seconds after reducing an initial 15-second pilot."
---

# DE Dost: Teaching Machines to Be a Friend on the Last Mile

[Swiggy](https://yomu.fyi/company/swiggy) · Aishwary Yadav · Jul 17, 2026

**Type:** Explainer

## Summary

DE Dost is a GenAI onboarding companion embedded in Swiggy’s delivery-executive app, designed for new DEs who receive information but lack confidence-building, immediate guidance. It supports questions in nine languages across SOPs, earnings, payouts, app navigation, shifts, and support, adapting detail and tone to conversational context. Before the first LLM call, a graph state assembles user context, preprocessing resolves references, and embedding-based semantic routing separates general knowledge-base questions from personal diagnostics. Retrieval uses authored QnA pairs plus hybrid semantic and lexical scoring to handle paraphrases and terms such as “MG” and “COD,” while guardrails constrain follow-ups to available knowledge. Evaluation spans about 800 cases per language across fleet and payout combinations, with macro accuracy trending at 90%+ and a target P50 latency of 4.5 seconds after reducing an initial 15-second pilot.

## Context

New delivery executives receive onboarding information through training and the DE app, but the information does not consistently build confidence or provide immediate answers. The source characterizes the underlying need as scalable mentorship: experienced DEs help newcomers prioritize, interpret procedures, and solve personal issues, but that support cannot reach everyone.

## Approach / What changed

DE Dost combines a conversational companion experience with a three-level personalization framework, graph-state context, query preprocessing, embedding-based routing, QnA-structured retrieval, and hybrid semantic-plus-lexical reranking. It adapts language and response detail, routes personal diagnostics to relevant data, and uses evaluation, feedback reviews, latency consolidation, and follow-up guardrails to improve quality and safety.

## Takeaways

- The system resolves conversational references during preprocessing, producing a normalized English query before retrieval; this prevents follow-ups such as “What if I miss it?” from being interpreted without their payout context.
- Semantic routing uses sentence-embedding similarity against seeded general and personal utterances, avoiding an extra LLM classification call and allowing the routing model to be updated by changing and re-embedding seed examples.
- The evaluation matrix covers about 200 questions across two fleet types and two payout types, producing roughly 800 test cases per language, with multilingual, adversarial, routing, and end-to-end response checks.

**Tags:** [Generative AI](https://yomu.fyi/topic/generative-ai), [Latency](https://yomu.fyi/topic/latency), [LLMs](https://yomu.fyi/topic/llm), [Retrieval-Augmented Generation](https://yomu.fyi/topic/retrieval-augmented-generation), [Testing](https://yomu.fyi/topic/testing)

- Source: [Swiggy](https://medium.com/swiggy-bytes/de-dost-teaching-machines-to-be-a-friend-on-the-last-mile-a4175e0f7066)
- Source URL: https://medium.com/swiggy-bytes/de-dost-teaching-machines-to-be-a-friend-on-the-last-mile-a4175e0f7066
- Ingested by Yomu: 2026-09-09T13:57:15.793Z

[Read original post](https://medium.com/swiggy-bytes/de-dost-teaching-machines-to-be-a-friend-on-the-last-mile-a4175e0f7066)