ak.Aman KumarFull-stack · AI · Cloud

Tab to navigate · Escape to close

Bengaluru, IndiaResume ↗
Find your way
Read the resume ↗

Aman Kumar / Software engineer

I build the
whole product.
From UI to AI
to cloud.

Full-stack engineering. Applied AI.
The infrastructure to bring it to life.

I understand the business, build the product, and own the details between the interface and production.

Explore my work Read my resume ↗
Full-stack at CodeAptoAWS Certified Solutions Architect
One product. Every layer.

React / Next.js / TypeScript

The part people actually use.

Customer websites, business dashboards and mobile workflows. Interfaces shaped around what people need to get done.

See the Procure-to-Pay platform

Tools I build with

ReactNext.jsTypeScriptRustPythonPostgreSQLAWSDocker

Selected engineering

Real problems.
Working systems.

From a shop’s daily billing to enterprise workflows and real-time communication. What I built, and the decisions behind it.

Production business software

Hardware Store ERP

One invoice. Every part in agreement.

Billing is never just a PDF. It changes stock, customer balances and the books. I built a retail and wholesale ERP that keeps those pieces together.

  • React
  • Node.js
  • PostgreSQL
  • Redis
  • AWS
Inside the system

Keep the transaction whole

Invoice creation uses a six-step PostgreSQL transaction with FOR UPDATE row locking. Invoice, stock and ledger updates succeed together, avoiding partial business records.

Move slow work out of the way

BullMQ and Redis workers generate PDFs with Puppeteer, then store them privately in S3. PostgreSQL pg_trgm search supports partial and misspelled product names.

Run it in the real world

Billing, inventory, pricing, unit conversion, customer ledgers and financial reports, deployed on Ubuntu with AWS RDS PostgreSQL and CloudFront.

Visit the platform ↗

Mar 2026 — present

Current engineering · CodeApto

Procure-to-Pay platform

A workflow is only as strong as its boundaries.

At CodeApto, I work across an enterprise procurement product: from purchase requests and supplier onboarding to approvals, sourcing and purchase-order release.

  • React
  • TypeScript
  • Rust / Axum
  • PostgreSQL
Inside the system

Connect the whole workflow

Building frontend features and backend flows across purchase requisitions, approvals, sourcing, supplier onboarding, award decisions and purchase-order release.

Make invalid transitions impossible

Improved workflow paths to prevent invalid status changes, duplicate actions and unauthorized PO release, with tests around the main paths.

Respect the boundary

Role-based access, tenant-specific data handling and workflow rules are part of the feature, not something added after the interface. This account uses only public resume-level information.

Jun 2026 — present

Vartalaap / message delivery

One message.
Independent deliveries.

AWS LambdaWebSocket APIDynamoDB

An illustrated fan-out

Message
AWaitingBWaitingCWaiting

Try sending with one recipient offline. Watch which deliveries complete.

A recipient going offline should not stop everyone else’s conversation.

Serverless communication

Vartalaap

Conversations move. Failures stay contained.

A real-time communication platform built around authenticated connections, clear data access patterns and message delivery that can tolerate an individual failure.

  • Next.js
  • Lambda
  • DynamoDB
  • Cognito
  • Chime SDK
Inside the system

A connection with an identity

Authenticated WebSocket connections support one-to-one and group messaging, with audio/video calls and push notifications using AWS-managed services.

Design around access patterns

DynamoDB single-table access patterns underpin the communication backend. Parallel fan-out with Promise.allSettled keeps one failed recipient from stopping the rest of the broadcast.

Carry the work into operations

My deployment experience also includes AWS EC2, Route 53, IAM and Security Groups at Technosys, with Docker and GitHub Actions workflows for repeatable releases and simpler rollback.

Independent project

Applied intelligence

AI, inside
the product.

Useful experiences built around language, voice and vision. Connected to the people and workflows they serve.

Shiv Cement Store product interface

Business platform

Shiv Cement Store

A retail platform with a multilingual RAG assistant, connected web and mobile tools, and real business workflows.

Next.js · Express · MongoDB · Expo · OpenAI · AWS S3
AI-Powered Alexa Assistant product interface

Voice / AI

AI-Powered Alexa Assistant

A serverless voice assistant connecting Alexa to OpenAI for contextual conversations, with session persistence in DynamoDB.

AWS Lambda · Node.js · OpenAI · DynamoDB
MedScan AI product interface

Vision / AI

MedScan AI

An image-and-text medication information tool, using Gemini Vision to identify medicines and present readable summaries.

Gemini Vision · AWS · Next.js · React Native

How I work

The whole product.
Not just my part.

Meet the engineer ↗
Understand

Start with the business.

The users, the daily friction, the constraints. A useful product begins with the right problem.

Build

Connect every layer.

Interface, API, data and AI where it helps. I follow the feature across boundaries.

Operate

Design for what comes next.

Transactions, background workers, caching and repeatable deployments. Concrete foundations for growth.

More things I’ve shipped

The project archive.

AI experiments, business platforms,
cloud infrastructure and focused tools.

8 / 8 projects
Business platform

Shiv Cement Store

2025 — 2026
Shiv Cement Store project screenshot

A customer website, staff/admin mobile app and REST API for a cement retailer. Product enquiries, quotations and staff workflows connect across web and mobile.

Includes a multilingual RAG assistant, GPS and rotating-QR attendance, role-based authentication and S3 uploads. Existing architecture artwork is retained.

Next.js · Express · MongoDB · Expo · OpenAI · AWS S3

Visit project ↗View architecture ↗Permalink
Voice / AI

AI-Powered Alexa Assistant

2025
AI-Powered Alexa Assistant project screenshot

A serverless voice assistant connecting Alexa to OpenAI for contextual conversations, with session persistence in DynamoDB.

Built with AWS Lambda and Node.js, with CloudWatch and secrets held in Systems Manager Parameter Store.

AWS Lambda · Node.js · OpenAI · DynamoDB

Visit project ↗Permalink
Creative tools

Mythos AI Studio

Archive
Mythos AI Studio project screenshot

An AI storytelling experiment that turns prompts into narrated videos through multi-agent orchestration.

Combines image generation and video composition in a Streamlit interface.

Python · Streamlit · Stable Diffusion · MoviePy

Visit project ↗Permalink
Commerce

Foxpop.in

Archive
Foxpop.in project screenshot

An e-commerce platform for makhana, with a product catalogue, secure payments and deployment pipelines.

Built on the MERN stack and deployed with AWS EC2 and Nginx, with Razorpay for payments.

MERN · AWS EC2 · Nginx · Razorpay

Visit project ↗Permalink
Vision / AI

MedScan AI

Archive
MedScan AI project screenshot

An image-and-text medication information tool, using Gemini Vision to identify medicines and present readable summaries.

A web and mobile project, also distributed as an APK. It is an information tool; its output is not a substitute for professional medical advice.

Gemini Vision · AWS · Next.js · React Native

Visit project ↗Permalink
Web utility

URL Shortener

Archive
URL Shortener project screenshot

A small, focused utility: turn a long URL into a short, shareable link.

Built with Next.js and TypeScript, deployed on Vercel.

Next.js · TypeScript · Vercel

Visit project ↗Permalink
Language / AI

Prompt Enhancer

Archive
Prompt Enhancer project screenshot

A utility for refining language-model prompts into clearer instructions.

A focused interface around the OpenAI API, built with Next.js.

Next.js · OpenAI · Vercel

Visit project ↗Permalink
Cloud infrastructure

Serverless Notification System

Archive
Serverless Notification System project screenshot

An event-driven architecture for email and SMS notifications, with Step Functions coordinating state and delivery logic.

A cost-conscious serverless project built with Lambda, API Gateway, SNS and SES. The original architecture image remains available here.

AWS Lambda · Step Functions · SNS · SES · Python

View architecture ↗Permalink

Engineering notes

Thinking through
the details.

Read the blog ↗

For teams with ambition. And problems worth solving.

Let’s build
something useful.

Start a conversation ↗amankeshri7479@gmail.com

राम Curiosity in the work. A quiet constant underneath.