Open to roles & freelanceDelhi NCR, India

Backend & AI
Engineer.

I’m Krrish Kumar. I build Node.js APIs, background jobs, and real-time features. At Parivar AI, I built the backend that lets you call an AI persona in Hindi and pick up a conversation later.

Krrish Kumar
Krrish Kumar, based in Delhi NCR, India.

Selected work

I built the backend for each of these projects. Pick one to see what happens after the user taps a button.

AI product · Co-founder

Parivar AI

A Hindi and Hinglish family companion. I built the backend for voice calls, persona conversations, and long-term memory.

Node.js · PostgreSQL · pgvector · LLM APIs

Parivar AI · how it's built

Website for Parivar AI (opens in a new tab)

Google Play · Currently being redeployed

Real-time backend

RideWave

I built a ride-matching service that finds drivers near a pickup with Redis and sends each one an offer over Socket.IO.

Redis GEO · Socket.IO · MongoDB

RideWave · how it's built

Source for RideWave (opens in a new tab)

Code execution platform

AlgoForge

I built a code judge that queues submissions in BullMQ, runs each one in a disposable Docker container, and sends the verdict back to the browser.

BullMQ · Docker · Fastify

AlgoForge · how it's built

Source for AlgoForge (opens in a new tab)

Trace a request

Pick a project, then step through one request from the user’s first action to the result.

Book a ride

Which drivers should see your request, and what happens to the other offers when someone accepts?

Read the RideWave case study
Express · MongoDB
Save the booking

You choose a pickup and destination. The API estimates the fare from the distance and saves a pending booking.

Open source

I publish small tools for problems I run into building backends. These are my two latest releases.

llm-pacer

When several workers call the same model, each sees only part of the bill. llm-pacer gives them one Redis-backed budget for requests, tokens, and concurrent calls.

Redis · LLM APIsSource

secret-blast-radius

An offline check for credentials exposed to npm install scripts. It lists what needs attention and suggests a rotation order while keeping secret values out of the report.

CLI · Dependency securitySource
Browse 6 more packages

webhook-crash-lab

Send the same webhook twice and check whether the customer was credited twice. It also tests deliveries that arrive late or out of order.

CLI · Webhook testingSource

webhook-crash-lab-mcp

Run the webhook test suite from an editor through MCP and compare the results after changing a handler.

MCP · Webhook testingSource

eslint-plugin-bullmq

Flags mistakes in BullMQ workers, including missing error listeners and job IDs that fail to prevent duplicate work.

ESLint · BullMQSource

create-bullmq-app

Starts a TypeScript queue service with an API, a worker, health checks, and graceful shutdown already wired up.

TypeScript · BullMQSource

zod-india

Validates Indian identifiers in Zod, including GSTIN, PAN, IFSC, and UPI IDs. It also formats rupee amounts in lakhs and crores.

Zod · ValidationSource

Merged upstream

Experience

  1. Dec 2025 – Present

    Launched on Google Play, being redeployed

    Parivar AI

    Co-founder & Backend/AI Engineer

    • Co-founded an AI companion app (100+ Google Play downloads) where users chat with, and get voice calls from, 9 AI family personas in Hindi, Hinglish, and English.
    • Built the Node.js backend (my co-founder built the Flutter client), with a multi-provider LLM layer: automatic Gemini → OpenAI → Groq fallback, per-attempt timeouts, and streamed replies.
    • Added long-term memory in PostgreSQL + pgvector with nightly consolidation, and real-time voice calls over WebRTC with a LiveKit agent (streaming STT → LLM → TTS).
    parivarai.in(opens in a new tab)
  2. Oct 2025 – Dec 2025

    Rablo

    Backend Developer Intern

    • Developed REST APIs in Node.js, Express, and MongoDB with JWT authentication, route authorization middleware, and standardized API error handling.
    • Refactored backend logic into controller and service layers to improve code organization and maintainability.
  3. 2022 – 2026

    B.Tech in Computer Science & Engineering

    Ajay Kumar Garg Engineering College, Ghaziabad, India

How I build

Three calls I made in my own projects, and what each one cost.

RideWave

Send the offer only to drivers near the pickup.

I keep driver locations in Redis and send the booking to the sockets that match a radius search. I also keep track of who got it, so I can take it off their screens when someone accepts. The part I’d change next is confirmation. Two drivers can still accept at nearly the same time.

Airline Booking System

The email can wait. The seat count can’t.

I keep the seat update in the booking request and put the reminder on RabbitMQ. A slow mail server can delay the receipt while the passenger already has a booking result. That leaves me with a separate reminder worker to run and monitor.

Parivar AI

Keep the voice call separate from the chat API.

My co-founder built the Flutter app. I put the speech-to-text, model reply, and speech playback in a LiveKit agent, and kept conversation memory in the backend. That gave us a separate place to work on call quality, at the cost of another service to deploy.

AI engineering

Voice and memory are already part of my production work. I’m now building retrieval pipelines and learning how to evaluate them.

  1. Shipped

    AI in a shipped product

    Parivar AI talks to users in Hindi, calls them in a natural voice, and reacts to the photos they send.

    LLM APIs · Prompt design · Conversation management · Text-to-speech · Image input

  2. Building

    AI grounded in your data

    Retrieval over your own documents, and structured outputs other services can consume.

    RAG · Embeddings · Vector search · Structured outputs · FastAPI · Tool calling

  3. Next

    AI you can trust in production

    Agents with bounded autonomy, measured by evaluations and held in place by guardrails.

    Agents · LangGraph · MCP · Evaluations · Guardrails · LLMOps

View full AI skill map
Production
Used in shipped products or public working systems
Hands-on
Actively building with it
Learning
Studying and experimenting

LLM applications

LLM features behind real APIs: typed inputs, typed outputs, streamed responses.

  • LLM API integration, Production
  • Prompt engineering, Production
  • Conversation management, Production
  • Gemini · OpenAI · Groq APIs, Production
  • Structured outputs · Pydantic, Hands-on
  • Function / tool calling, Hands-on
  • FastAPI services, Hands-on
  • Streaming · SSE, Production
  • Async AI workflows, Hands-on

RAG systems

Retrieval pipelines that ground answers in sources and can prove it.

  • Embeddings, Production
  • Semantic search, Production
  • ChromaDB, Hands-on
  • LangChain, Hands-on
  • Retrieval pipelines, Hands-on
  • Metadata filtering, Learning
  • Hybrid search · BM25, Learning
  • Cross-encoder reranking, Learning
  • Citations · grounding, Learning
  • LlamaIndex, Learning
  • Pinecone, Learning
  • RAG evaluation · RAGAS, Learning

AI agents

Stateful, tool-using workflows with explicit control flow and bounded autonomy.

  • Tool-using agents, Hands-on
  • Memory · context management, Production
  • LangGraph, Learning
  • Stateful workflows, Learning
  • MCP servers and tools, Learning
  • Multi-agent systems, Learning
  • CrewAI · AutoGen · smolagents, Learning

Multimodal & voice

Speech and vision as first-class inputs and outputs.

  • Text-to-speech, Production
  • Image input to LLMs, Production
  • Speech-to-text, Production
  • Voice agents, Production
  • CLIP · computer vision, Learning

Production AI

The backend half of AI: evaluation, guardrails, observability, cost, and latency.

  • Retries · fallbacks, Production
  • Output validation, Hands-on
  • Latency · cost optimisation, Learning
  • Prompt caching, Learning
  • Evaluations · LLM-as-judge, Learning
  • Guardrails · prompt-injection defence, Learning
  • Observability (Langfuse, LangSmith), Learning
  • Local LLMs · Ollama, Learning
  • LLMOps, Learning

Deep learning foundations

Enough of the model side to reason about what the API is doing.

  • PyTorch · training loops, Learning
  • CNNs · batch normalisation, Learning
  • RNNs · LSTMs · GRUs, Learning
  • Transformer architecture, Learning
  • Tokenisation · BPE, Learning
  • Fine-tuning, Learning

Not claimed: ML research or training models at scale. The deep learning layer is there so I understand what the model is doing behind the API.

Technical toolkit

The tools I’ve shipped with and the ones I’ve used in hands-on projects.

Production
Used in shipped products or public working systems
Hands-on
Actively building with it

Backend

  • Node.js, Production, used in Every backend I have built
  • Express.js, Production, used in RideWave, Airline, AlgoForge, Rablo
  • Fastify, Production, used in AlgoForge Submission Service
  • REST API design, Production, used in All backend work
  • Microservices, Production, used in Airline Booking, AlgoForge
  • API gateway, Production, used in Airline API Gateway
  • JWT auth · bcrypt, Production, used in Auth Service, RideWave, Rablo
  • Socket.IO, Production, used in RideWave, AlgoForge Socket Service
  • Rate limiting, Production, used in Airline API Gateway
  • Schema validation (Zod), Production, used in AlgoForge Evaluator
  • FastAPI, Hands-on, used in AI engineering projects

Data

  • MongoDB · Mongoose, Production, used in RideWave, AlgoForge, Rablo
  • MySQL · Sequelize, Production, used in Airline Booking (five services)
  • Redis, Production, used in GEO in RideWave, queues and socket maps in AlgoForge
  • PostgreSQL, Production, used in Parivar AI (self-hosted on AWS EC2)
  • pgvector, Production, used in Parivar AI long-term memory
  • Prisma, Production, used in Parivar AI
  • SQL, Production, used in Airline Booking, Parivar AI
  • Vector databases (ChromaDB), Hands-on, used in RAG projects

Distributed systems

  • RabbitMQ, Production, used in Airline Booking → Reminder Service
  • BullMQ, Production, used in AlgoForge submission and evaluation queues
  • Event-driven messaging, Production, used in Airline Booking, AlgoForge
  • Async job pipelines, Production, used in AlgoForge
  • Geospatial queries, Production, used in RideWave (Redis GEORADIUS)
  • Sandboxed execution, Production, used in AlgoForge (Docker per submission)

AI engineering

  • LLM API integration, Production, used in Parivar AI
  • Prompt engineering, Production, used in Parivar AI personas
  • Conversation management, Production, used in Parivar AI
  • Text-to-speech, Production, used in Parivar AI voice calls
  • Image input to LLMs, Production, used in Parivar AI photo reactions
  • Embeddings, Production, used in Parivar AI memory (pgvector)
  • Retrieval pipelines, Hands-on
  • Structured outputs · Pydantic, Hands-on
  • Function / tool calling, Hands-on
  • FastAPI services, Hands-on

Infrastructure

  • Docker · dockerode, Production, used in AlgoForge Evaluator
  • Structured logging (Winston), Production, used in AlgoForge
  • Queue monitoring (Bull Board), Production, used in AlgoForge
  • Git · Postman, Production, used in All work
  • AWS EC2 · PM2, Production, used in Parivar AI
  • Nginx, Production, used in Parivar AI reverse proxy
  • AWS S3, Hands-on, used in Personal projects
Product surface
Next.js · React · Tailwind CSSShipped on this site
Languages
JavaScript · TypeScript · Python (hands-on) · C++

About me

Portrait of Krrish Kumar
AKGEC · CSE 2022–2026 · Delhi NCR

I like figuring out what happens after someone taps a button.

I got into backend work building a flight booking system. Getting one passenger a seat and an email took five services. Later, on RideWave, I had a different problem to solve. Several drivers could see the same offer, but only one should get the ride.

In December 2025 I co-founded Parivar AI. My co-founder built the Android app, and I built the service behind the conversations. These days I’m interested in how to tell whether an AI remembered the right detail from an earlier chat.

Building
Parivar AIBeing redeployed
Shipping
Open-source dev tools8 npm packages · upstream fixes
Studying
Distributed systemsMIT 6.824 · DDIA
Open to
Backend & AI rolesIncluding founding-engineer seats

Explore in the terminal

Try projects to find a case study, packages for my npm releases, or prs to read a change that made it upstream.

Have a system to build?

I'm open to backend and AI engineering roles, founding-engineer opportunities, and freelance work.

Based
Delhi NCR, India
Timezone
IST, UTC+5:30

Leave a message

Krrish Kumar