BS Artificial Intelligence · Roblox developer

AI.Games.Software.Built with purpose.

BS Artificial Intelligence student and Roblox developer building practical AI tools, interactive games, and software that ships.

Explore selected workView GitHub
Scroll to see the work

01 · Published game

Roblox / Lua / Live

Escape Tsunami
with Dragons

A survival experience built around one clear pressure: keep moving, earn rewards, and survive the wave.

Escape Tsunami with Dragons Roblox experience thumbnail
30K+visits
Play on Roblox
What shipped

Tsunami rounds, lucky-block rewards, collectible dragons, music, tools, announcements, and progression systems.

The hard part

Making a large moving hazard feel smooth while protecting server performance and validating player actions.

What it taught me

Shipping matters. Performance, security, player behaviour, and iteration become real only after people play.

Development timeline

  1. 01

    Prototype

    Built the core survival loop around a moving tsunami, map traversal, and repeatable rounds.

  2. 02

    Systems

    Added lucky-block rewards, collectible dragons, tools, music, and player-facing progression.

  3. 03

    Performance

    Reworked movement and update frequency so the large hazard could feel smooth without wasting server time.

  4. 04

    Protection

    Added server-side validation, anti-exploit checks, and controlled admin announcements.

  5. 05

    Publish & iterate

    Shipped the experience, watched player behaviour, and kept improving the loop after release.

Open the live experience

02 · AI learning product

Team project · Streamlit / Python

Study Buddy

Static course files become a grounded study workflow: understand, practise, focus, and measure progress.

Built withAbdul Rehman · Adeel Mughal · Asad MehmoodCreateX AI App Development project
study_buddy / workspaceAI study system
0102030405

PDF · MD · TXT · PPTX

Upload academic material

lecture-06.pdfoperating-systems.pptxnotes.md
Product-flow visualization. Replace with real Study Buddy screenshots when available.
01

Upload academic material

Process PDF, Markdown, TXT, and PPTX files instead of forcing students to reorganize everything manually.

PDF · MD · TXT · PPTX
02

Extract grounded knowledge

Build context from the uploaded material and answer questions against that source to reduce unsupported responses.

Contextual document chat
03

Turn content into study tools

Generate structured notes, active-recall flashcards, and adaptive multiple-choice quizzes from the same material.

Notes · Cards · Quizzes
04

Sequence the learning path

Rank documents into a logical roadmap so foundational topics come before advanced concepts.

AI-ranked roadmap
05

Track work, focus, and progress

Combine a priority task manager, Pomodoro sessions, quiz analytics, and progress visualizations in one workflow.

Tasks · Focus · Analytics
View the repository

03 · Local-first AI application

FastAPI / Keras / SQLite

MindfulMirror

Emotion detection becomes more useful when it leads somewhere: a history, patterns, reflection, and a private next step.

127.0.0.1:8000Local-first
Webcam processed locally
DetectedFocused87%
Last 7 entries
Calm Focused Low
Reflection

What changed before your mood improved today?

Wellness assistant

Let’s break the next step into something manageable.

OpenRouter · key stored locally
Interface visualization based on the real feature set. Add app screenshots in public/media for a fully literal showcase.
01

Camera

Capture webcam frames inside the browser while keeping the app local-first.

02

Emotion detection

Send frames to a local FastAPI backend and classify expressions with a local Keras model.

03

History

Save emotion entries and notes in a local SQLite database rather than a remote account.

04

Patterns

Turn the saved history into charts, counts, and recent mood activity.

05

Reflection

Offer optional Gemini-powered questions, advice, and insights with rule-based fallbacks.

06

Chatbot

Connect an OpenRouter-powered wellness assistant while keeping secrets in a local environment file.

Privacy is an architectural choice. Camera frames, emotion history, and API secrets are designed to stay on the user’s machine.

Explore MindfulMirror

04 · NLP / RAG experiment

Nutrition answers
grounded in source material.

A retrieval-augmented system built to answer nutrition questions from a controlled knowledge source instead of relying only on a model’s memory.

QuestionRetrieveGroundAnswer
Watch the LinkedIn demo

Proof of participation

Certificates that support the work — not replace it.

A small, relevant set: technical participation and professional development. Click any certificate to inspect it fullscreen.

Working stack

Tools I have used to make real systems.

No percentages. No made-up mastery bars. Just the technologies that appear in the work above.

01

AI & machine learning

PYPython
TETensorFlow
KEKeras
OPOpenCV
RARAG
GEGemini
OPOpenRouter
02

Applications & backend

FAFastAPI
STStreamlit
SQSQLite
HTHTML
CSCSS
JAJavaScript
03

Game development

LULua
RORoblox Studio
GAGame systems
PEPerformance
SEServer validation
04

Workflow

GIGit
GIGitHub
VSVS Code
RIRider
DODocumentation
RARapid prototyping

About

A student who prefers shipping over pretending.

I’m studying Artificial Intelligence at UMT and building across AI applications, Roblox game systems, and practical software projects. The common thread is simple: take an idea far enough that somebody else can actually use it.

I’m currently looking for internships where I can learn from a real team, improve my engineering judgement, and contribute work that reaches users.

Based in
Sialkot, Pakistan
Education
BS Artificial Intelligence, UMT
Focus
AI products · games · software