Khadija Fassih.
AI · Full-Stack · Android — Lahore, Pakistan

I build at every layer of the stack.

From a storage engine written in C++ that manages its own 4KB disk pages, up to agentic AI systems and the Next.js and native Android apps people actually touch. Computer Science Student at UET Lahore, and seven projects built to find out how things really work underneath.

View work Get in touch
Khadija Fassih
FocusAgentic AI, RAG pipelines, and full-stack product work.
Also buildsNative Android in Kotlin, and systems software in C++.
Core stackPython · TypeScript · Next.js 15 · FastAPI · PostgreSQL
StudyingBS Computer Science, UET Lahore — since 2023.
Selected work

Three projects, three different layers.

One that ships to users, one that goes all the way down to the disk, and one that had to solve a context problem no prompt could fix.

01

Repo-Sieve

Live · Full-stack context tool

What it does

Turns a live GitHub repository into clean, LLM-ready XML. You point it at a branch, pick what matters, and it hands back exactly the context a model needs instead of the whole tree.

What's notable

It stays responsive on trees of 100,000+ files, using a recursive UI and Zustand for constant-time lookups, with concurrent branch-aware fetching and automatic binary detection. Access is locked down with GitHub OAuth 2.0, HttpOnly JWT sessions, Cloudflare Turnstile, and Postgres-backed sliding-window rate limiting.

Next.js 15TypeScriptServer ActionsPostgreSQLPrismaZustandGitHub APIOAuth 2.0
Visit site ↗
02

LiteDB++

Relational database engine, from scratch

What it does

A working RDBMS written in C++ with nothing underneath it. Its own page-based storage engine, its own pager managing 4KB blocks, and a query path that runs tokenizer to parser to execution.

What's notable

The B-tree handles node splitting and redistribution properly, so lookups stay logarithmic under heavy inserts rather than degrading. Above it sits an LRU buffer pool with dirty-page management for ACID behaviour, and a later branch adds concurrency control on top.

C++Systems programmingB-tree indexingBuffer poolFile I/OConcurrency control
View code ↗
03

EchoTube

RAG over long-form video

What it does

Ask questions about a three-hour YouTube video and get an answer that links straight to the moment it came from. Retrieval-augmented generation is what breaks the context wall — the transcript never has to fit in the prompt.

What's notable

The ingestion pipeline chunks transcripts, embeds them with text-embedding-3-small, and batch-upserts them into Qdrant with timestamps attached, which is what makes the deep links possible. Answers stream over SSE from a LangChain LCEL chain served by FastAPI, retrieving the top five ranked chunks per question.

PythonFastAPILangChain (LCEL)RAGQdrantSupabaseSSE streamingReact + ViteTypeScript
Also built

Four more, across web and mobile.

Live · Agentic ERP

LabSphere

An AI-enabled lab management system with an agentic layer built on MCP servers, cutting the manual coordination and report compilation that eats a lab's week. Includes a team-formation engine that balances groups on skill mapping and live workload, delivered across eight Agile sprints with role-based access control.

MEAN stackMCPAirtable APINLPRBAC
Visit site ↗
Native Android

FocusQuest

A gamified productivity app in Kotlin, built on MVVM with five REST services behind one repository layer. A hybrid foreground service keeps the countdown alive through configuration changes and navigation, and the AI task planner uses a two-phase prompt protocol with exponential-backoff retries when the model misbehaves.

KotlinMVVMRoomRetrofitFirebaseGroq LLaMA
View code ↗
Cross-platform

Islamic Habit Tracker

A Flutter app for prayers and daily habits, wired to a prayer-times API with its own time-format conversion. Tracking resets itself on date rollover and aggregates progress across screens, with scheduled local notifications for prayer times and randomised habit reminders.

FlutterDartSQLiteLocal notifications
Watch demo ↗
Research project

Parental Digital Safety Platform

An Android platform for supervised-device monitoring and usage analytics, built on the Accessibility and Notification APIs. The backend handles encrypted data exchange and remote reporting, with device registration and sessions secured through JWT, encrypted local storage and token-refresh flows.

KotlinAndroid system APIsFastAPIReactPostgreSQLJWT
Experience

Professional development.

April 2026

Automation Developer

Datacom — AI Accelerator Program · Virtual internship simulation
  • Advised a client on automating the timesheet-to-invoice process end to end, applying agentic AI and workflow orchestration to remove manual bottlenecks.
  • Designed a roadmap for moving from co-pilot to autonomous agents, mapping automation maturity levels and picking out the highest-return processes.
  • Produced a stakeholder-ready proposal covering solution architecture, tool selection, risk assessment and projected efficiency gains.
Certificate ↗
Education

BS Computer Science

University of Engineering and Technology, Lahore
Sep 2023 — Present
Toolkit

What I work with.

Languages

  • Python
  • C++
  • Kotlin
  • TypeScript
  • JavaScript
  • SQL

AI engineering

  • LangChain
  • LangGraph
  • RAG
  • Agentic AI
  • MCP
  • Vercel AI SDK

Web

  • Next.js 15
  • React
  • Node.js
  • Express.js
  • FastAPI

Mobile

  • Android SDK
  • MVVM
  • Room
  • Retrofit
  • Coroutines & Flow
  • Firebase
  • Flutter / Dart
  • Material Design

Data

  • PostgreSQL
  • MongoDB
  • SQLite
  • Qdrant
  • Prisma
  • Supabase
  • Neon

Cloud & practice

  • Docker
  • Vercel
  • AWS
  • GitHub Actions
  • Git
  • Agile / Scrum
  • Systems design
  • DSA
Contact

Open to internships and junior developer roles.

Remote or Lahore-based. If you're building something with AI at the centre of it, or something that needs to work properly at the layer underneath, I'd like to hear about it.

Emailkhadijafassih87@gmail.com WhatsApp+92 324 6462350 GitHubgithub.com/khadijafassih LinkedInin/khadija-fassih