Marwa's Portfolio

about.me · Marwa

Hi I'm Marwa

// CS Student · AI Systems & Strategy · Dubai, UAE

CS student focused on applied AI: ML pipelines, NLP systems, and reinforcement learning. I like building things end-to-end and figuring out how they actually work. Open to collaborating on interesting AI projects.

LinkedIn GitHub

Recent Role

AI Developer @ StudentVerse

Education

CS (AI & Big Data) @ UOW Dubai · WAM 82

Location

Dubai, UAE

Focus

AI · ML · Big Data

projects · click to explore

02 · AI SYSTEMS

BrainBrew AI

github.com/Marwakhot/BrainBrew

PROBLEM

Passive PDF learning not engaging students in critical thinking.

SOLUTION

AI-powered platform using NLP to generate adaptive questions and difficulty levels that promote deeper understanding.

ReactTypeScriptAI/MLNLPSupabasePostgreSQLThree.js

// ARCHITECTURE

User uploads PDF

↓

React + Three.js Frontend

↓

NLP Engine

• Text extraction & chunking

• Adaptive question generation

• Difficulty level scoring

↓

Supabase + PostgreSQL

↓

Interactive Q&A Session

01 · AI SYSTEMS

ORION Talent AI

github.com/Marwakhot/Orion

PROBLEM

Manual hiring processes are slow, biased, and lack structured onboarding for new employees.

SOLUTION

End-to-end HR platform with AI CV matching, behavioral assessments, a LangChain + RAG chatbot, and automated onboarding.

ReactTypeScriptNode.jsLangChainRAGAI/MLPostgreSQL

// ARCHITECTURE

Recruiter

Candidate

↓

React + TypeScript Frontend

↓

Node.js REST API

↓

AI Layer

• CV parsing → skill matching

• LangChain + RAG chatbot

• Behavioral voice/text assessment

• Composite scoring engine

↓

PostgreSQL + pgvector

↓

Shortlist → Onboarding → Day One Plan

01 · MACHINE LEARNING / NLP

AutoBug · Bug Triage System

github.com/Marwakhot/bug-triage-nlp-api

PROBLEM

Slow bug assignment causing delays in development cycles and inefficient resource allocation.

SOLUTION

NLP + ML system to automatically predict the responsible team and priority for each bug report, with bulk predictions via a Streamlit dashboard.

PythonNLP (TF-IDF)LightGBMFlask APIStreamlitDockerInternal ToolingWorkflow Automation

// ARCHITECTURE

Bug Report Input

↓

NLP Pipeline

• TF-IDF vectorization

• Text preprocessing

↓

LightGBM Classifier

• Team prediction

• Priority prediction

↓

Flask API

Streamlit Dashboard

↓

Docker Container

03 · AI SYSTEMS

NeuroNav · Robot Navigation AI

github.com/Marwakhot/Gazebo_robot_navigation

PROBLEM

Autonomous navigation challenges in complex environments with dynamic obstacles.

SOLUTION

Implemented Fuzzy Logic, Behavior Trees, and Q-Learning for obstacle avoidance in a 10m x 10m simulated arena.

PythonFuzzy LogicBehavior TreesReinforcement Learning

// ARCHITECTURE

Gazebo Simulation Environment

↓

Sensor Input

• Obstacle detection

• Position mapping

↓

Decision Engine

• Fuzzy Logic controller

• Behavior Trees

• Q-Learning (RL)

↓

Robot Actions → 10m x 10m Arena

02 · MACHINE LEARNING / NLP

LoanGuard · AI Risk Management Audit

github.com/Marwakhot/-AI-Risk-Management-Audit

PROBLEM

AI loan-approval systems can carry hidden bias and security gaps that go undetected without formal auditing.

SOLUTION

A fairness and security audit using the NIST AI RMF: quantified gender and age bias with the 80% Rule, found a 20% manipulation success rate on the age field, and delivered prioritized fixes.

PythonScikit-learnFairlearnPandasMatplotlibNIST AI RMFJupyter

// ARCHITECTURE

Loan Application Data

↓

NIST AI RMF Framework

• GOVERN: scope & objectives

• MAP: risk identification

• MEASURE: quantify bias

• MANAGE: recommendations

↓

Fairlearn + Scikit-learn

↓

Audit Report + Visualizations

03 · MACHINE LEARNING / NLP

QuantRisk · VaR Validation & Capital Allocation

github.com/Marwakhot/QuantRisk-VaR-Validation

PROBLEM

Investment portfolios need statistically validated risk estimates to set accurate capital reserves.

SOLUTION

Built a three-method VaR framework (Historical Simulation, Variance-Covariance, Monte Carlo) with Kupiec's POF backtesting. All three passed validation, converging on a $21,578 daily capital reserve at 99% confidence.

PythonPandasNumPySciPyMonte CarloyfinanceJupyter

// ARCHITECTURE

yfinance → 3yr Historical Prices

• AAPL, MSFT, JPM, GLD, SPY

↓

01 · Data Setup

• Normality tests (Jarque-Bera, Shapiro-Wilk)

• Correlation mapping

↓

02 · VaR Calculation (3 Models)

• Historical Simulation

• Variance-Covariance (Parametric)

• Monte Carlo (10,000 scenarios)

↓

03 · Backtesting

• 70/30 train-test split

• Kupiec's POF statistical test

↓

Capital Reserve = $21,578 @ 99% confidence

hackathon · click to explore

Space42 Bounty Challenge · 3rd Place

Google Developers Group · Univ. of Birmingham Dubai · Nov 2024

ORION · AI-Powered Talent & Onboarding Partner

Won 3rd place building ORION in a single weekend: an end-to-end HR platform with AI-powered CV matching, behavioral assessments, and automated onboarding workflows.

CV upload with instant eligibility matching against job skills
Behavioral AI assessment through voice or text
Combined scoring system for smart candidate shortlisting
Conversational AI chatbot powered by LangChain and RAG
Employee zone unlock with structured Day One plan

ReactTypeScriptNode.jsLangChainRAGAI/ML

Marwa and teammates at the Space42 Bounty Challenge, Google Developers Group, University of Birmingham Dubai

experience & achievements · click to explore

// CURRENT

AI Developer

StudentVerse

Research Trainee

UOWD Tech Club

CS Student

UOW Dubai

// EXPERIENCE

Data Science Intern

Alamar

AI Rep · Gitex 25

Dreamloop.ai

Head Girl

NIMS Dubai

Director General

NIMSMUN

MUN Chairperson

Multiple Conferences

High School Grad

NIMS Dubai

AI Developer

StudentVerse · May 2026 – Aug 2026

Built the backend for SV Orbit, a multi-mode AI chatbot platform live on the App Store, including Semantic Search.

Built the backend for Semantic Search in SV Orbit, one of 2 shipped modes at the time
Integrated LLM APIs and custom logic to take the mode from backend to production

Research Trainee

UOWD Tech Club · May 2026 – Present

Researching Spiking Neural Networks and neuromorphic AI in the Tech Club's Research Department.

Reviewing academic literature on neuromorphic architectures and biologically-inspired computing
Evaluating how these methods could integrate into conventional AI pipelines
Contributing structured literature reviews to the department

Computer Science Student

UOW Dubai · Major: AI & Big Data · WAM 82

Computer Science degree specializing in AI & Big Data, WAM 82.

Coursework spanning ML systems, data engineering, and applied AI
Building hands-on skills through independent projects in ML, NLP, and data engineering

Data Science Intern

Alamar Foods · Dunkin' · Dubai · Oct 2025 – Apr 2026

Built data pipelines and automated dashboards for Dunkin' operations, turning 2M+ rows of raw data into decision-ready insights.

Engineered AWS Redshift pipelines processing 2M+ rows, writing SQL to produce clean, analysis-ready datasets
Built KPI dashboards tracking sales pipeline, customer acquisition, and revenue trends
Reported metrics to cross-functional leadership on daily, weekly, and monthly cadences

AI Product & Solutions Ambassador

Dreamloop.ai · GITEX Global 2025 / Expand North Star

Represented Dreamloop.ai at GITEX 2025, presenting AI product capabilities to 60+ enterprise clients.

Engaged 60+ enterprise clients, including CEOs and founders, explaining AI product value to non-technical audiences
Delivered live demos of text-to-game and text-to-app generation
Represented the brand at one of the world's largest tech events

Head Girl

NIMS Dubai

Led the student council and served as the main representative between students and school administration.

Organized 15–20 in-school and inter-school events end-to-end, coordinating with external sponsors, vendors, and community partners to secure funding
Represented a 5,000+ student body, communicating student needs to administration and external parties
Coordinated a 35-member council to deliver events drawing 200–300 attendees each

Director General · Model United Nations

NIMSMUN · NIMS Dubai

Organized and led a full 3-day MUN conference with 150+ delegates from across the UAE.

Led external outreach to 20+ schools across the region, coordinating with faculty advisors and student delegations to recruit 150+ participants
Secured sponsorships and served as the central point of contact across a 50-member organizing team, guest speakers, and participating schools
Delivered the full conference end-to-end across committees, sessions, and logistics

MUN Chairperson

Multiple MUN Conferences

Chaired committees at multiple Model United Nations conferences, guiding delegates through structured diplomatic discussions.

Moderated debates and maintained professional procedure
Ensured fair participation across all delegates
Helped delegates develop diplomacy, research, and public speaking skills

High School Graduate · Science Stream

NIMS Dubai

Completed high school with a focus on science and analytical subjects, building the foundation that led to pursuing Computer Science.

Built a strong foundation in logical thinking and problem solving
Developed early interest in technology and data
Active in extracurriculars and academic competitions