Skip to main content
Portrait of Avinash Narwariya

About

How I got here, from a voice assistant in tenth grade to AI systems. ↓

  1. Tenth grade

    The first spark

    In tenth grade I started building my own version of Jarvis in Python. I wired speech to text, text to speech and a few voice services together until it could listen to me and answer back.

    Watching code I wrote actually do something is what got me interested in programming.

  2. Twelfth grade

    Curiosity returns

    In twelfth grade I built a management system in Python, connected to a database, for the computer science subject in my curriculum.

    It brought the interest back, and this time I wanted to explore programming more seriously.

  3. College

    Starting with DSA

    When I came to college I started learning data structures and algorithms.

    After three or four months I realized that solving isolated problems wasn't what excited me most. I was more of a builder.

    I wanted to create things, see results right away, and turn ideas into working products.

  4. College

    Web development clicked

    So I started learning web development, and it clicked. Writing code and seeing the result in a browser a second later made programming much more exciting.

    I got deeply interested in building interactive applications.

  5. The last year and a half

    Becoming a full stack developer

    Over the past year and a half I have kept learning and building projects, moving from frontend development to backend engineering, APIs, databases, application architecture and deployment.

    I wanted to understand how to build and ship an entire product, beyond the interface.

  6. Then

    Discovering generative AI

    My curiosity led me to generative AI, LLMs, embeddings and RAG.

    I began exploring how to put intelligence into software and build applications that could do more than conventional programs.

  7. Then

    From AI features to AI systems

    I became more and more interested in AI agents, tool use, orchestration, memory and reliable AI workflows.

    Integrating a model into an application stopped being enough. I wanted to understand how to build complete AI systems that could do useful work.

  8. August 2025

    Shipping with a team

    My first role was at eéna, as the founding backend engineer. I built media and reel microservices with Node.js and Cloudinary, an adaptive feed service with Redis caching, and containerized the backend with Docker.

    After that I kept working, building products with other teams. Those roles are below.

Today I build AI software by combining full stack engineering with AI systems, and I keep exploring how to turn complex ideas into useful, reliable products.

Here are the places I've worked

  1. to

    Kakiyo

    Full Stack EngineerRemote

    • Built the backend of the support platform used by Kakiyo, designing APIs and data models that power ticketing and customer conversations end to end.
    • Shipped multiple CRM integrations for Kakiyo, syncing contacts, deals, and activity across third-party CRMs so teams work from a single source of truth.
  2. to

    The Housekraft

    Full Stack DeveloperOnsite

    • Integrated a full-stack feedback system so users can submit, route, and track input alongside existing product flows.
    • Shipped Three.js-powered 3D rendering features, iterating on scene setup and performance so visual experiences stay responsive in the browser.
    • Improved onboarding UX by restructuring the flow, tightening copy and steps, and reducing friction before users reach core product value.
  3. to

    AptiDude

    Full Stack DeveloperRemote

    • Built end-to-end Puzzles and Blogs modules for a production platform, connecting a React user interface to backend APIs.
    • Designed and built an SEO-optimized blog CMS and admin dashboard with reusable frontend components and server-side optimizations for content discoverability and performance.
    • Developed a rating-based puzzle system with backend services and optimized database queries to handle concurrent users and keep the application stable.
  4. to

    eéna

    Founding Backend EngineerRemote

    • Developed media and reel microservices with Node.js and Cloudinary, including automated video processing pipelines for transcoding, streaming, and thumbnail generation.
    • Implemented an adaptive feed service with infinite scrolling, Redis-based caching, and rule-based ranking logic to improve content delivery efficiency and user engagement.
    • Containerized backend microservices using Docker and built an internal analytics backend to track user interactions and platform metrics, so the team could make data-driven feature improvements.

What I work with

How I build

Coding agents are a big part of how I work. I set up the harness around them, the instructions, tools and checks, so they can carry a task further on their own and I spend my time on the decisions.

  • Claude Code
  • Codex
  • Cursor
  • OpenCode
  • Pi
Frontend
  • HTML
  • CSS
  • JavaScript
  • TypeScript
  • React
  • Next.js
  • Tailwind CSS
  • shadcn/ui
  • Framer Motion
  • TanStack Query
  • Three.js
Backend
  • Node.js
  • Bun
  • Express.js
  • Fastify
  • Hono
  • Socket.IO
  • FastAPI
  • Python
  • Spring Boot
  • Zod
Data
  • MongoDB
  • PostgreSQL
  • TimescaleDB
  • Prisma
  • Drizzle
  • Redis
  • Kafka
AI & Agents
  • LangChain
  • LangGraph
  • Gemini
  • OpenRouter
  • MCP
Infra & Tooling
  • Docker
  • Nginx
  • AWS
  • Vercel
  • Cloudflare
  • Railway
  • Turborepo
  • Temporal
  • Prometheus
  • Grafana
  • Puppeteer
  • GitHub
  • Postman

Have a role in mind?

Send me a message on X with the role and a link to the posting. My resume has the rest.

Send me a messageBook a call
Building a product instead? See how I work with founders.