SDE 3 AT ZOMATOBACKEND × PRODUCTION AI

Rishav Kumar. Backend engineer building production AI.

IN PRODUCTION

Engineering the systems
behind useful intelligence.

From merchant platforms to AI that works.

EXPLORE THE WORK

01 / PRODUCTION AICURRENT WORK · ZOMATO

Intelligence,
connected.

A merchant asks a question.
The answer depends on a whole system.
I lead engineers building that system.

A QUESTION BECOMES A WORKFLOWINTERACTIVE ARCHITECTURE SKETCH
Merchant AI architectureConversation and durable context connect to native model orchestration, business tools and knowledge retrieval, followed by evaluation. This is an illustrative flow, not a live product. Merchant contextUSER · RESTAURANT · MEMORY ConversationRESPONSE-ID CONTINUITY Nativeorchestration Knowledge retrievalBusiness toolsFILE SEARCH + REUSABLE APIS Correctness · language · tool execution Merchant context and conversation connect to native orchestration, knowledge retrieval, business tools and evaluation. Merchant contextUser · restaurant · memory ConversationResponse-ID chaining Nativeorchestration File searchBusiness APIsKnowledge retrievalReusable tools Evaluation & observability
01 / THE RIGHT CONTEXT

Conversation history and durable merchant context have different jobs. Response-ID chaining preserves the conversation; cross-session memory carries the user and restaurant context.

Illustrative architecture, simplified to show the engineering decisions. No live business data.

MY WORK

The model is one
part of the product.

I lead merchant AI engineering across orchestration, integrations, context and the tools that help teams operate the system. Recent work includes rebuilding prompt-heavy flows around the OpenAI Responses API, native tool calling and file search.

Reusable APIs and stable tool schemas make new workflows easier to integrate. Evaluation and observability make the behavior inspectable—from a failed tool call to a response in the wrong language.

Responses API / native tools / file searchCross-session memory / evaluation / observability

02 / BACKEND SYSTEMSGO · MERCHANT PLATFORMS

A stronger
foundation.

Before the assistant,
the platform.
Connecting the work of running a business.

LEGACY PHP → GO SERVICES

One onboarding journey.
Many connected systems.

I led migration of merchant onboarding from a legacy PHP service into Go, bringing verification, payments and commission plans into a clearer service workflow.

01 / VERIFICATION

Image detection and document checks connect to the onboarding flow, so identity and account setup can move through the same service.

LESS REPEATED WORK

Built to be reused.

Onboarding replication for multi-brand kitchens. Partner-facing flows for location changes, ownership transfers and access invitations.

OPERATIONS INCLUDED

Built to be understood.

Monitoring, priority alerts and operational tooling support the service after launch. My broader work includes merchant commissions and point-of-sale order migration.

03 / APPLIED MACHINE LEARNINGDATA SCIENCE · 2022–2023

From model
to decision.

My work started with prediction.
Then with the systems that put it to use.

LIGHTGBM + MLFLOW

A prediction has to
reach the product.

Conceptual model development and deployment illustration, not measurement dataDEVELOPEVALUATEDEPLOY

Developed and deployed real-time recommendation models with LightGBM and MLflow, alongside rule-based strategies. The work connected model behavior with product and commercial decisions.

Conceptual illustration · no performance data
THE THROUGH LINE

Models.
Workflows.
Production.

Working across data science and backend engineering changed how I approach AI: deployment, evaluation and product behavior belong in the same conversation.

Python · LightGBM · MLflow
Prophet · ARIMA

04 / THE FULL PICTURERISHAV KUMAR

A path through
the whole system.

I'm an SDE 3 at Zomato, where I lead engineers building merchant AI and the services behind it. My work spans LLM orchestration, tool execution, retrieval, memory and evaluation, alongside Go-based merchant platforms.

I started in applied machine learning after studying Mathematics at IIT Kanpur, with a minor in Algorithms.

I care about clear interfaces, reliable behavior and measurable product outcomes.

Connect on LinkedIn
THE PROGRESSION

From mathematics
to models,
to the systems
around them.

  1. 2026-07 — Present

    SDE 3

    Zomato

  2. 2025-06 — 2026-07

    SDE 2

    Zomato

  3. 2023-06 — 2025-06

    SDE 1

    Zomato

  4. 2022-04 — 2023-06

    Data Scientist

    Zomato

  5. 2021-12 — 2022-04

    Data Science Intern

    Zomato

  6. 2021-05 — 2021-07

    Summer Intern

    Walmart Global Tech India

  7. 2018-09 — 2021-04

    Team Leader

    VISiON IITK

  8. 2018 — 2022

    Mathematics

    Indian Institute of Technology Kanpur

HOW I WORK
01

Make the behavior visible.

Monitoring, alerts and evaluation help a team see where a system needs attention.

02

Teach the fundamentals.

I've mentored engineers through system design, networking, databases and web fundamentals.

03

Connect the disciplines.

Model behavior, service design and product requirements are parts of the same engineering problem.

TOOLS I HAVE WORKED WITH

BackendGo · Python · PHP · APIs · Gin

Data & infrastructureMySQL · DynamoDB · Redis · Kafka · SQS · AWS · Docker

AI & MLResponses API · Native tools · File search · LightGBM · MLflow · Prophet · ARIMA

ObservabilityNew Relic · Datadog · Grafana · Prometheus