Krrish Kumar

Backend engineer working in Node.js and TypeScript, moving deliberately into production AI engineering. I have built a real-time ride-matching backend on Redis GEO and Socket.IO, a five-service airline system coordinated through RabbitMQ, and a four-service code-execution platform that runs untrusted submissions in throwaway Docker containers. I co-founded Parivar AI, an AI companion app launched on Google Play (currently being redeployed), and built its production backend: multi-provider LLM chat, vector memory in PostgreSQL and pgvector, and real-time voice calls, deployed on AWS.

1. Co-founder & Backend/AI Engineer · Parivar AI

   Dec 2025 – Present

   India · Launched on Google Play, being redeployed

   • 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).
   • Deployed on AWS EC2 (PM2, Nginx) with self-hosted PostgreSQL and LiveKit.

2. Backend Developer Intern · Rablo

   Oct 2025 – Dec 2025

   Remote

   • 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.

1. Parivar AI

   AI product · Co-founder

   Millions of Indians live away from family, in hostels, new cities, or abroad. Parivar AI gives them an AI family member to talk to when they cannot call home: a Maa, Papa, or Dadi persona that replies in Hindi and Hinglish, calls back in a natural Hindi voice, reacts to photos, and remembers what matters.

   AI behind a service boundary · Language as a product requirement · Honest by design. Launched on Google Play with 100+ downloads; currently being redeployed. I built the Node.js backend, deployed on AWS EC2 (PM2, Nginx) with self-hosted PostgreSQL and LiveKit; my co-founder built the Flutter client.

   Node.js · Express.js · PostgreSQL · pgvector · Prisma · Gemini · OpenAI · Groq · LiveKit · AWS EC2

   Visit Parivar AI

2. RideWave

   Real-time backend

   Match a passenger to nearby drivers the moment a ride is requested, push the offer to those drivers instantly, and withdraw it from everyone else once one of them accepts, without polling the database.

   Redis GEO for proximity · Addressable sockets, not broadcasts · Monolith on purpose. A working real-time matching loop: request, proximity search, targeted offer, confirm, and withdrawal from the drivers who lost the race.

   Node.js · Express · Socket.IO · Redis GEO · MongoDB · Mongoose · JWT

   View source

3. Airline Booking System

   Microservices

   Split an airline backend into services with clear ownership, authenticate at the edge, and keep booking fast even though every booking triggers email that depends on a slow, unreliable SMTP server.

   Service boundaries follow ownership · Asynchronous where the slow dependency is · Authentication at the edge. A five-service system with a single entry point, per-service data ownership, and booking latency decoupled from email delivery.

   Node.js · Express · MySQL · Sequelize · RabbitMQ · JWT · bcrypt · Nodemailer

   View source

4. AlgoForge

   Code execution platform

   Run code you did not write, in four languages, on your own hardware, without letting it escape, without blocking the API, and deliver the verdict to the user the moment it exists.

   Queues on both sides of execution · A container per submission · Push, don't poll. End-to-end judge pipeline across four independently deployable services, with queue monitoring through Bull Board and structured logs through Winston.

   Node.js · TypeScript · Fastify · Express · BullMQ · Redis · Docker · MongoDB · Socket.IO

   View source

Production

Used in shipped products or public working systems

Hands-on

Actively building with it

Learning

Studying and experimenting

Production

Node.js · Express.js · Fastify · REST API design · Microservices · API gateway · JWT auth · bcrypt · Socket.IO · Rate limiting · Schema validation (Zod) · MongoDB · Mongoose · MySQL · Sequelize · Redis · PostgreSQL · pgvector · Prisma · SQL · RabbitMQ · BullMQ · Event-driven messaging · Async job pipelines · Geospatial queries · Sandboxed execution · Docker · dockerode · Structured logging (Winston) · Queue monitoring (Bull Board) · Git · Postman · AWS EC2 · PM2 · Nginx · JavaScript · TypeScript · C++ · Next.js · React · Tailwind CSS · LLM API integration · Prompt engineering · Conversation management · Gemini · OpenAI · Groq APIs · Streaming · SSE · Embeddings · Semantic search · Memory · context management · Text-to-speech · Image input to LLMs · Speech-to-text · Voice agents · Retries · fallbacks

Hands-on

FastAPI · Vector databases (ChromaDB) · AWS S3 · Python · Structured outputs · Pydantic · Function / tool calling · FastAPI services · Async AI workflows · ChromaDB · LangChain · Retrieval pipelines · Tool-using agents · Output validation

Learning

Pinecone · Metadata filtering · Hybrid search · BM25 · Cross-encoder reranking · Citations · grounding · LlamaIndex · RAG evaluation · RAGAS · LangGraph · Stateful workflows · MCP servers and tools · Multi-agent systems · CrewAI · AutoGen · smolagents · CLIP · computer vision · Latency · cost optimisation · Prompt caching · Evaluations · LLM-as-judge · Guardrails · prompt-injection defence · Observability (Langfuse, LangSmith) · Local LLMs · Ollama · LLMOps · PyTorch · training loops · CNNs · batch normalisation · RNNs · LSTMs · GRUs · Transformer architecture · Tokenisation · BPE · Fine-tuning

Ajay Kumar Garg Engineering College, Ghaziabad, India