DOC AT-2026-001 STATUS Under active review
PRODUCT REQUIREMENTS DOCUMENT · CANDIDATE SPEC

Aditya Tripathi — Full Stack Developer & AI/ML Engineer

A specification for what I've built, how I build it, and what I'm proposing to build next — written the same way I write a PRD before I write a line of code.

OPEN TO
INTERNSHIPS
Owner
Aditya Tripathi
Version
v3.2
Last updated
Aug 2026
Status
Approved for review
01

Problem Statement

Most CS students graduate with a resume full of claims and a GitHub full of tutorials. There's a gap between "I know React" and "I shipped a system that 500 real people depend on" — and recruiters can't always tell the difference from a bullet point.

This document exists to close that gap: every claim below is backed by a live URL, a published package, or a repo you can actually open.

Note I'd rather show you one system that's actually live than five that only run on localhost.
02

Proposed Solution

I'm a third-year B.Tech CSE student at Feroze Gandhi Institute of Engineering & Technology, Lucknow (CGPA 8.1/10). My working method is deliberate, not accidental: PRD → technical architecture → DB schema → API contract → UI spec → tickets, before a single line of implementation gets written.

It's slower on day one and faster on every day after — it's why the systems I ship keep running instead of quietly breaking three weeks later. I'm currently deepening my DSA fundamentals through the NeetCode 150, optimizing for real understanding over memorized patterns.

Long-term, I'm building toward founding something myself — I'd rather solve a real problem end-to-end than optimize someone else's dashboard forever. Right now, that means an internship where I can ship production code alongside people better than me.

Note This is also just how I think — spec first, code second. Ask me about any doc and I can walk you through the reasoning.
03

Scope — Shipped Modules

MODULE 03.1 · nirashrayfoundation.com
Nirashray Foundation
Shipped · 500+ users
Next.jsTypeScriptPostgreSQLPrismaNextAuth v5Razorpay

A full NGO management system for a cow rescue & welfare organization — public site, member portal, and an admin CMS/ERP dashboard, built as 1.17M+ lines of TypeScript.

  • RBAC via NextAuth.js v5 — Admin, Member, Guest roles with session management
  • Razorpay donation processing with live transaction tracking & a financial analytics dashboard
  • Auto-generated ID cards, receipts & certificates with QR-code validation (jsPDF)
  • Tiptap-based CMS with Cloudinary uploads and real-time content preview
Note This one started as a favor for an NGO that needed a proper system, not another spreadsheet. It's now the project I'm proudest of — 500+ members actually use it every week.
MODULE 03.2 · github.com/aditya1046785
AI Resume Generator
RAG pipeline
Next.jsTypeScriptGoogle GeminiVector EmbeddingsGitHub OAuth

An end-to-end RAG system that matches your real GitHub projects to a job description via cosine similarity, then writes ATS-ready bullet points from the best matches.

  • GitHub OAuth pulls repos and auto-summarizes each via Gemini
  • Embedding pipeline + cosine similarity surfaces top-fit projects with relevance scores
  • Resume writer agent turns matches into quantified, ATS-optimized bullet points
  • JSON → HTML → PDF export via headless browser rendering
MODULE 03.3 · resume-craft-ai-rouge.vercel.app
ResumeCraft AI
Shipped
Next.jsFirebase GenKitGeminiZustandZod

A conversational resume builder — Gemini asks the questions, validates answers in real time, and assembles the resume section by section with a live preview.

  • Context-aware question engine, validated with Zod + React Hook Form
  • AI rewrites weak bullet points into stronger, quantified phrasing
  • "Fit Score" compares resume content against a job description semantically
  • Zustand-synced live preview with one-click PDF export
MODULE 03.4 · https://www.npmjs.com/package/ai-metrics
AI-Metrics — LLM Usage SDK
Published package
JavaScriptNode.jsSDKTelemetry

A production-ready npm package that tracks LLM API usage — token counts, cost estimation, and privacy-safe telemetry — for OpenAI and Anthropic calls.

  • Configurable async event batching (time- or size-based flush)
  • Exponential backoff with jitter for reliable event delivery
  • Local cost-estimation engine — no external API calls required
  • Redacts API keys and secrets before transmission (GDPR-minded)
MODULE 03.5 · https://github.com/aditya1046785/framesense-app
FrameSense
Real-time framework
TypeScriptNext.jsWebSocketsJWTGenKit

A real-time AI interface framework built for sub-second latency, with a full JWT auth system and modular AI orchestration via GenKit.

  • Bidirectional WebSocket layer for sub-second interactions
  • Token generation, validation middleware & refresh-token rotation
  • RESTful endpoints for users, AI interactions & data sync
  • Standardized prompt templates for modular AI feature integration
04

Dependencies

Languages

  • typescript
  • javascript
  • python
  • sql

Frontend

  • next.js ^16
  • react ^19
  • tailwindcss
  • zustand

Backend

  • node.js / express
  • rest & websockets
  • next-auth ^5
  • prisma orm

Databases

  • postgresql
  • mongodb
  • firebase
  • vector-db

AI / ML

  • google-gemini-api
  • firebase-genkit
  • rag-pipelines
  • vector-embeddings

Tooling

  • git / github
  • vercel
  • razorpay-sdk
  • zod
05

DSA Proficiency

I'm building this deliberately, not by grinding random problems — tier by tier, working through the NeetCode 150 for structured coverage, prioritizing conceptual understanding over memorized patterns. Started at fundamentals, now pushing into advanced technique.

Note Numbers below are pulled straight from my LeetCode profile — this one updates as I go, not a one-time snapshot.

Fundamental

  • array x13
  • string x11
  • two-pointers x7

Intermediate

  • hash-table x12
  • sliding-window x5
  • greedy x2

Advanced

  • dynamic-programming x3
  • divide-and-conquer x2
  • quickselect x1
06

Success Metrics

1.17M+
Lines of TypeScript shipped
500+
Active NGO members served
2
Live production deployments
99.9%
Uptime on Vercel
07

Rollout Plan

Next step: a conversation, not another form field.

If this spec matches what you're hiring for — SDE, Full Stack, or AI/ML internship — I'm ready to start. Email is fastest.

Note Based in Lucknow, open to remote roles. I usually reply within a day.