01 — Portfolio · 2026

Jesseman
Devamirtham.

Full-stack engineer in Chennai. I design and ship complete production systems — React and Next.js front ends, Node and FastAPI back ends, PostgreSQL data layers, auth from primitives, and Web3 wallet flows. Front to back, schema to deploy.

Currently
B.Tech CSE · VIT Chennai
Final year
02 — About

I design and ship production systems end to end.

Not CRUD wrappers — auth servers built from primitives, API gateways with rate-limited token introspection, real-time data pipelines that hold up in prod. The unglamorous parts most portfolios skip: data integrity, edge cases, deploy reliability.

Stack discipline: TypeScript on Next.js for the front end, Node or FastAPI for the back, PostgreSQL with Prisma for the schema layer, Redis for state. JWT, OAuth 2.0, RBAC done right. Web3 via ethers.js and MetaMask. Containerized deploys.

AI-native by default. Published one IEEE Xplore paper during undergrad on geospatial contact tracing with ball-tree indexing. Capstone in progress on Vision Transformer activation functions. PyTorch + Hugging Face + LLM APIs as default tooling, not buzzwords.

Final-year Computer Science at VIT Chennai. Open to full-time roles starting mid-2026.

03 — Stack

The tools, and how I use them.

The full stack — front to back. No skill bars, no percentages. Just what I reach for in production, with a line on why.

01

Web

Production web applications, end to end. App Router patterns, type-safe data flow, Web3 wallet integration, edge runtimes.

Daily
  • TypeScript
  • React
  • Next.js (App Router)
Shipped
  • Tailwind CSS
  • Framer Motion
  • ethers.js / MetaMask
  • Radix UI
  • shadcn/ui
02

Backend

Schema-first APIs. REST and GraphQL where each fits. Strong opinions on error handling and observability.

Daily
  • Node.js
  • Express
  • Python / FastAPI
Shipped
  • REST
  • GraphQL
  • WebSockets
  • Webhook handlers
03

Data

Schema design that won't bite you in month 6. Indexes where they matter, denormalize where it pays.

Daily
  • PostgreSQL
  • MongoDB
  • Redis
Shipped
  • Prisma ORM
  • Mongoose
  • Drizzle
  • Pinecone (vector)
04

Auth & security

Auth servers from primitives. JWT and OAuth 2.0 done right. Rate limiting in the right layer.

Daily
  • OAuth 2.0
  • JWT
  • NextAuth.js
Shipped
  • RBAC
  • ABAC
  • API gateway design
  • Rate limiting
05

AI / ML

Research-grade PyTorch. LLM APIs for production. ViT, MARL, adversarial training as default frameworks.

Daily
  • PyTorch
  • LLM APIs (Claude, OpenAI)
  • Hugging Face
Shipped
  • Vision Transformers
  • Multi-agent RL
  • Gradio
  • LangChain
06

Infrastructure

Containerized deploys to Vercel and beyond. CI that actually catches things. Observability before it's a fire.

Daily
  • Docker
  • Vercel
  • GitHub Actions
Shipped
  • Linux
  • Sentry
  • Microservices
  • Edge functions
04

Selected work.

Five builds across Web3, real-time interfaces, and production frontends — wallet integration, server-driven state, sub-second loads. Each one shipped.

01Web3 · dApp

PollChain

Decentralized voting dApp with MetaMask wallet auth, live on-chain vote counting, and a no-wallet demo mode. Six routes, animated real-time results.

Problem

Web3 UX is the wall most users hit — wallet connection, transaction states, and on-chain latency break the interface if the frontend isn't built for them.

Solution

Built the full frontend in Next.js 14 + TypeScript: ethers.js wallet integration, React Context for chain state, animated progress bars driven by live vote data, and a demo fallback so anyone can explore without a wallet.

Stack
  • Next.js 14
  • React
  • TypeScript
  • ethers.js
  • MetaMask
  • Tailwind
02Interactive UI

TaskFlow

Full-stack Kanban board with native drag-and-drop, multi-board management, and type-safe server-driven mutations. Strict-mode TypeScript throughout.

Problem

Drag-and-drop boards get janky fast — optimistic UI, persistence, and re-render control fight each other once state grows.

Solution

Built on Next.js 14 App Router with Server Actions for every mutation, native HTML5 drag-and-drop, a priority/label/due-date task model, search + filter, and Prisma migrations. No client-state library — server actions keep it type-safe end to end.

Stack
  • Next.js 14
  • TypeScript (strict)
  • Server Actions
  • Prisma
  • SQLite
  • Tailwind
03Performance

NexusFlow AI

High-performance marketing + onboarding site for an AI workflow platform. Lighthouse 98 / 100 / 100 / 100, custom animation system, type-safe content layer.

Problem

Client needed an enterprise-grade, founder-facing site that loads instantly and holds up under scrutiny — shipped in under two weeks.

Solution

Next.js App Router with a type-safe content layer, a hand-built Framer Motion animation system, and ICP-tuned copy. Tuned to a near-perfect Lighthouse score across all four axes.

Stack
  • Next.js
  • TypeScript
  • Tailwind
  • Framer Motion
  • Vercel
04Dashboards · Payments

VitalEdge Fitness

Coaching SaaS with client dashboards, programming, and Stripe payment flows. Built solo, end to end — the premium-user systems a token economy needs, minus the chain.

Problem

Client ran the whole business on DMs, Google Docs, and Stripe payment links. No dashboard, no billing logic, no source of truth.

Solution

Built the full product: typed dashboards for clients and programs, Stripe Connect split payments, scheduled reminder jobs, and a Prisma schema that won't bite in month six.

Stack
  • Next.js
  • TypeScript
  • PostgreSQL
  • Prisma
  • Stripe
  • Vercel
05Product

Kaval

Shipped production frontend for an AI digital-safety agent that sits between the user and the internet. Built on the latest Next.js, live in production.

Problem

A new AI product needs a frontend that explains an unfamiliar concept fast and converts — on a bleeding-edge Next.js with breaking API changes.

Solution

Built the production site on Next.js 16 App Router, reading the framework's own docs to ship against APIs newer than most tooling supports. Clean, fast, conversion-tuned.

Stack
  • Next.js 16
  • React
  • TypeScript
  • Tailwind
  • Vercel
05 — Track record

What I've been building.

  1. 2026 — Present

    Founder· Stratum

    Custom AI operating systems for solo operators — Claude Code workspaces with skills, commands, hooks, and automations, governed by one config. One-time builds across four tiers.

  2. Summer 2025

    Intern — Digital ME · Hyundai Motor India

    Data analysis + error correction on 590+ robotic systems across the production line. Python, SQL, IIoT data pipelines.

  3. 2024 — Present

    Independent Engineer · Client work

    Built and shipped 4 client SaaS platforms (NexusFlow AI, Solara Studio, VitalEdge Fitness, Artisan Eats). End-to-end from schema design through Vercel deployment.

  4. 2022 — 2026

    B.Tech Computer Science · VIT Chennai

    Final year. Focus on systems, distributed computing, and applied ML. Published one paper in IEEE Xplore during undergrad on geospatial contact tracing with ball-tree indexing. Capstone in progress on Vision Transformer activation functions.

06 — Contact

Building something
production-grade?

Open to full-time engineering roles starting mid-2026, and selective client builds in the interim. The fastest path is email.