Skip to content
const rahul = {
  role: "Full Stack Dev",
  experience: "3 yrs",
  stack: ["React", "Next", "Node"],
  loves: "AI + clean UX",
};
 
ship(rahul);

Portfolio of Rahul Bisht, Full Stack Developer

Rahul Bisht

  • Full Stack Developer
  • AI & LLM Apps
  • 3+ years
scroll

Hi! Nice to meet you

My name is Rahul Bisht

  • Full Stack Developer
  • AI & LLM Apps

I'm a full stack developer with 3 years of experience building scalable web applications with React, Next.js, Node.js, Python and LangChain.

I care about software that is efficient, secure and maintainable — from accessible, internationalised UIs to Backend-for-Frontend layers and RAG-powered AI workflows. I'm always exploring what's next in AI and modern web development.

Education

2019 – 2023

Bachelor of Technology

Mahatma Gandhi Mission's College of Engineering & Technology

Noida, NCR · GPA 8.13

Daily drivers

  • Re
  • Nx
  • No
  • Ts
  • Py
  • Lc

Experience

  1. to

    Full Stack Developer

    Telgoo5 Inc

    A group of Vcare Technology Inc & Telgoo5 Inc

    Next.js, Node.js & MongoDB apps with SSR/SSG, a BFF layer, and RAG workflows on AWS Bedrock monitored with LangSmith.

  2. to

    Full Stack Developer

    Seventh Triangle Consulting

    React, Python & Node.js apps; WCAG-compliant, fully internationalised RTL UIs; GraphQL & REST integrations.

Portrait of Rahul Bisht
Rahul BishtIndiain/rahul-bisht

Interests

  • AI Agents
  • RAG & LLM Evals
  • Web Performance
  • Accessibility
  • System Design
  • Developer Tooling

Selected work

01. Payments & BFF

HBC Telecom

Client Project · Telecom e-commerce

A telecom storefront where customers browse plans and check out securely. I integrated PCI-DSS-compliant Stripe payments with tokenised card data, so card details never touch our servers.

A Backend-for-Frontend aggregates downstream APIs and keeps credentials server-side, while a Zustand cart re-validates prices on the server and DOMPurify keeps every input XSS-safe.

Tools

  • Nx
  • Ts
  • No
  • St

Skills

  • BFF architecture
  • Payments
  • Security
  • AI orchestration
mobile.hbci.com/plans
HBC.mobile

Pick a plan that moves with you

Starter

5 GB

$25/mo

Add to cart
Popular

Unlimited

∞

$45/mo

Add to cart

Plus

15 GB

$35/mo

Add to cart

One request in, many APIs out

  1. Browser — 1 request per view
  2. BFF layer — aggregates & keeps tokens server-side
  3. Auth service
  4. Plans API
  5. Stripe
  6. AI services
  • Tokenised cards

    PCI-DSS Stripe flow — raw card data never touches app servers.

  • Server-side pricing

    Zustand cart, but prices are re-validated on the server at checkout.

  • XSS-safe inputs

    DOMPurify sanitisation and strict validation on every field.

  • Secrets stay server-side

    AI & partner credentials live behind the BFF, never in the bundle.

02. AI Recommendations

Question 2 of 4

What's your style?

  • Minimal
  • Bold & bright
  • Classic
  • Sporty
Next →

Just In Time

Client Project · AI product discovery

A quiz-based product discovery flow. I built the Node.js recommendation backend with Brotli/gzip compression and HTTP caching, combining quiz answers with product context for AI-assisted recommendations.

On the front end, React quiz filters use code splitting, lazy loading and memoised rendering to improve Core Web Vitals — plus an AI-assisted theme migration with Cursor that removed unused CSS/JS.

Tools

  • Re
  • No
  • Js
  • Cu

Skills

  • Recommendations
  • Caching
  • Core Web Vitals
  • AI-assisted dev

From answers to recommendations

  1. 1

    Quiz answers

    Structured responses from React filters

  2. 2

    Product context

    Catalogue data merged with answers

  3. 3

    AI ranking

    Assisted recommendation workflow

  4. 4

    Fast delivery

    Brotli/gzip + HTTP caching

  5. 5

    Results

    Relevant picks, low latency

Core Web Vitals, tuned

Code splitting, lazy loading and memoised rendering on the quiz filters, plus an AI-assisted theme migration that stripped unused CSS/JS.

LCPfaster first paint
INPsnappier filters
Lighthouseleaner bundle

03. Accessibility & i18n

FNP Qatar

Client Project · Gifting storefront

I independently built the complete front-end architecture in Liquid, HTML, CSS and JavaScript — fully responsive, from concept to production.

The UX is WCAG-compliant and i18n-ready: ARIA roles, screen-reader and keyboard support, RTL layouts with CSS logical properties, locale-aware validation, and GraphQL/REST with Accept-Language negotiation.

Tools

  • Lq
  • Js
  • Gq
  • Cs

Skills

  • Accessibility
  • RTL & i18n
  • GraphQL
  • UI components
fnp.qa
fnpSearch 'Anniversary Gifts'
  • Premium
  • Gift Sets
  • Birthday
  • Anniversary
  • Flowers
  • Cakes
  • Plants
  • Personalised

Blooms & Bakes

Fresh flowers and cakes, made to turn moments into celebrations.

Explore Now

Shop By Occasion

  • Birthday
  • Anniversary
  • New Born
  • Love
Delivery toDohaع
  • All gifts
  • Birthday
  • Anniversary
Search 'Anniversary Gifts'

FlowersShop all

  • Blush Rose BouquetQAR 135now45–60 min
  • Birthday Bloom BoxQAR 350Same Day

English · LTR

Built on the FNP brand kit

One token set drives both locales — colours, spacing and components stay identical while direction and typeface switch with lang and dir.

  • Brand olive#727500
  • Gift cream#f8f3ea
  • Ink navy#141c2b
  • Express orange#f7931e

Type pairing

Aa

DM Sans

English · LTR

أب

تجوال

Tajawal · RTL

Flowers & Gifts·زهور وهدايا

Accessible in every language

  • ARIA roles & landmarks
  • Screen-reader support
  • Full keyboard navigation
  • RTL via dir + logical properties
  • :lang-scoped typography
  • hreflang & x-default alternates
  • Locale-aware form validation
  • Accept-Language + UTF-8 APIs
Navigate withTabShiftEnterEsc— no mouse needed.
card.cssRTL-ready
.card {
  padding-inline-start: 1rem;
  border-inline-end: 2px solid;
  text-align: start;
}

:lang(ar) .card {
  font-family: "Tajawal";
}

<!-- locale-aware SEO -->
<link rel="alternate"
      hreflang="ar-QA" />

04. AI & RAG

RAG Workflows

Vcare · Production AI

RAG-based workflows built with LangChain, AWS Bedrock, vector search and LLM APIs — retrieving relevant context before generation to ground responses for knowledge-driven use cases.

Evaluated and monitored with LangSmith: response quality, latency, token usage and failure cases tracked across workflows to support reliable production deployments.

Tools

  • Lc
  • Py
  • Aw
  • Ls

Skills

  • RAG
  • LLM evaluation
  • Observability
  • Prompt engineering
assistant.internal / knowledge-chat
Which documents do I need to activate a prepaid plan?

Assistant

You'll need a valid government ID and proof of address. Activation completes after verification, usually within minutes.

activation-policy.pdf · p.3kyc-guide.md grounded
Ask anything…

answers cite their sources

Retrieval-augmented generation, end to end

  1. Query

    User question

  2. Embed

    Vectorise the query

  3. Retrieve

    Vector search, top-k

  4. Context

    Relevant chunks

  5. Generate

    LLM on AWS Bedrock

  6. Answer

    Grounded + cited

Evaluated & observed with LangSmith

Illustrative view of the signals I monitor across AI workflows.

Response quality

  • Grounded
  • Partially grounded
  • Needs review

Latency (p95)

Token usage

Failure cases

  • Retriever timeouttraced
  • Low-relevance contexttraced
  • Ungrounded claimtraced

05. Impact in Numbers

By the Numbers

Results · Vcare

Measured outcomes from my current role, where I own rendering strategy, a Backend-for-Frontend layer and production AI workflows.

Faster pages, fewer round-trips and leaner orchestration — while keeping tokens and secrets server-side.

Tools

  • Nx
  • No
  • Ts

Skills

  • SSR/SSG
  • BFF
  • Performance
The journey so far
B.Tech · MGM COET
Seventh Triangle
Vcare
'19'20'21'22'23'24'25'26'27
  • 35%

    faster initial page load

    SSR/SSG with Next.js

  • +18

    Lighthouse performance points

    Rendering & asset optimisation

  • 45%

    fewer client-side network requests

    Backend-for-Frontend layer

  • 30%

    less API orchestration overhead

    Aggregated downstream APIs

Before vs. after

Relative to the original baseline (100%). Shorter bars are better.

less is more

  • Client-side network requests

    before
    55%
  • API orchestration overhead

    before
    70%
  • Initial page load time

    before
    65%

06. Special Selection

The Toolbox

Special selection

Everything I reach for, grouped the way I think about building products — from the first pixel to the last trace in production.

currently exploring

  • LangGraph agents
  • FastMCP servers
  • OpenTelemetry for LLMs

Frontend

  • JavaScript (ES6+)
  • TypeScript
  • React
  • Next.js
  • HTML5
  • CSS3
  • Zustand
  • Liquid

Backend

  • Node.js
  • Express.js
  • Python
  • FastAPI
  • GraphQL
  • REST
  • BFF

AI

  • LangChain
  • LangGraph
  • LangSmith
  • AWS Bedrock
  • RAG
  • AI Agents
  • FastMCP
  • Prompt Engineering

Data & Ops

  • MongoDB
  • PostgreSQL
  • OpenTelemetry
  • Stripe
  • Git

Craft

  • Accessibility (WCAG/ARIA)
  • i18n & RTL
  • UI/UX
  • Postman
  • Jira
  • Cursor AI
  • Project Management

Thank you!

Thanks for taking the time to explore my work. I'm open to new roles, freelance projects or simply a good conversation about the web and AI.
Shall we connect?