Fatima Bilal Hassan

Fatima Bilal Hassan

CS at LUMS | AI & RAG Specialist | Full-Stack Developer building for scale and social impact

Based in: Lahore, PakistanMost recently: Data Analysis Intern, Japan International Cooperation Agency

Hire me

CS graduate from LUMS with a strong foundation in AI and full-stack development, specializing in RAG pipelines and transformer-based models. Proven track record of building high-impact platforms, from AI-powered news aggregators to collaborative architecture tools, with a deep focus on solving real-world problems like gendered micro-aggression detection and educational accessibility.

Experience

Data Analysis Intern

Japan International Cooperation Agencyinternship2026 – Mar 2026

Summary

Performed data cleaning, preprocessing, and EDA on education and socioeconomic datasets to support evidence-based policy recommendations. Applied statistical analysis and data visualization to identify trends, generating insights and visual summaries that informed project decision-making. Contributed to analytical reports and presentations delivered to project stakeholders to support data-driven planning.

Results

  • Authored the analysis and visual summaries for the national Non-Formal Education (NFE) Report 2024–25.
  • Produced the 'What Keeps Children Out of School' report, identifying that a mother's education and employment were significantly associated with child enrollment.
  • Conducted descriptive statistics, crosstabs with chi-square tests, and t-tests on national survey data to identify drivers of school dropout rates.
  • Analyzed annual data tables covering over 30,000 NFE centres, 34,000 teachers, and 1.2 million learners.
  • Processed enrollment trends dating back to 2019–20.
  • Analyzed a national survey dataset of 500 respondents across 10 districts and Islamabad.
Statistical AnalysisData VisualizationEDAExcelPython

Skills

Technical

PythonPython
NLP
RAG pipelines
LangChain
FlutterFlutter
Node.jsNode.js
Machine Learning
scikit-learn
Streamlit
transformer-based models
Hugging Face
Ollama
MongoDBMongoDB
GitGit
REST APIs
SQLSQL
JavaScriptJavaScript
ReactReact
AWS S3
PyTorchPyTorch
PL/SQLPL/SQL
FastAPIFastAPI
DockerDocker
VercelVercel
DartDart
NginxNginx
Statistical Analysis
Caching
Tool calling
Structured output parsing
Excel
Quantization
Hybrid Search

Projects

Gendered Micro-Aggressions Detection

Developeracademic
Built text classification pipelines using TF-IDF + Logistic Regression and XLM-RoBERTa on ~20,000 samples to detect explicit and implicit gendered microaggressions, improving accuracy from 88.35% to 92.6% with the transformer model. Mitigated severe class imbalance (11,822 non-sexist vs. 1,809 sexist examples) through oversampling and class weighting, significantly boosting minority class recall for subtle harmful content detection. Extended to a cross-lingual Urdu setting using Helsinki-NLP translation; TF-IDF baseline achieved 81.5% accuracy while transformer-based cross-lingual transfer revealed instability (73%), highlighting generalization challenges for low-resource languages.
TF-IDFLogistic RegressionXLM-RoBERTaHelsinki-NLPPython

O-Level Vault - AI-Powered Exam Prep Platform

Full-stack Developerpersonal
Identified a gap in O-Level exam resources and built a full-stack web platform using React where students can browse past papers and attempt AI-generated quizzes tailored by subject and topic. Built web scraping pipelines to collect and structure past-paper data, then trained a Hugging Face transformer model for dynamic quiz generation - removing dependence on manual content curation. Designed role-based access control for students, parents, and admins, enabling personalized progress tracking and structured content management for admins.
ReactWeb scrapingHugging Face transformerFull-stack

ArchConnect - Collaborative Architecture Platform

Developerpersonal
Developed a real-time collaboration platform for architects, clients, and vendors using Flutter, Node.js, and MongoDB, containerized with Docker Compose and deployed on Vercel - streamlining coordination across roles. Designed AWS S3 document storage with folder-level visibility inheritance, so each stakeholder only sees what they need - securing sensitive project files without manual permission management. Implemented Google OAuth and REST API authentication alongside a real-time notification system, enabling secure login across user types and timely updates throughout the multi-user platform.
FlutterNode.jsMongoDBDocker ComposeVercelAWS S3Google OAuthREST API

HCI User Research & Project

Researcheracademic
Synthesized user interviews and secondary research into design requirements for an assistive application supporting autistic children and their families.
HCIUser Research

AI-Powered News Summarization & Retrieval System

Developerpersonal
Addressed information overload by building an AI-powered news platform using Python, Streamlit, Ollama, and Hugging Face embeddings that aggregates articles from multiple sources and generates concise, unified event summaries. Built an interactive RAG pipeline with semantic search, vector embeddings, cosine-similarity retrieval, and clustering, enabling users to query daily news and receive context-grounded responses. Implemented real-time RSS ingestion, article deduplication, caching, background refresh, and top-story detection to ensure users always see fresh and relevant news.
PythonStreamlitOllamaHugging Face embeddingsRAGSemantic searchVector embeddingsCosine-similarityClusteringRSS ingestionLlama 3.2 3BBAAI/bge-large-en-v1.5

Electricity Billing System

personal
A three-tier application deployed on Oracle Cloud using a FastAPI server, Nginx, and Oracle DB to manage complex billing and payment processing.
SQLPL/SQLFastAPIOracle CloudNginxOracle DB

LLM-Powered RAG Chatbot with LangChain & Agentic Workflows

Developerpersonal
Designed a LangChain-based application using LCEL with modular pipelines for summarization, sentiment analysis, and report generation - enabling a single system to handle diverse NLP tasks with task-specific processing. Implemented and compared standard RAG, RAG-Fusion, and dynamic chain routing to address irrelevant context retrieval, improving response relevance by directing queries to the most appropriate retrieval strategy. Extended the system into a fully interactive assistant by developing an agent-powered chatbot with tool/function calling and a Streamlit interface, enabling real-time responses, external API usage, and memory across a conversation.
LangChainLCELRAGRAG-FusionDynamic chain routingStreamlitAgentic WorkflowsTool callingJSON schemas

Education

BSc. in Computer Science

Lahore University of Management SciencesComputer ScienceSep 2022 – May 2026

Coursework

AIDatabasesSoftware EngineeringAdvanced ProgrammingHCIData ScienceNLPNetwork Security

A-Level / O-Level

Beaconhouse, Gujranwala2017 – 2022

Recognition

Awards

Merit Scholarship 25% (O levels)

Beaconhouse School system

Merit Scholarship 100% (A levels)

Beaconhouse School system
Hire FatimaReach out about a role, a contract or a conversation.For recruiters

Fatima's twin is AI, it can make mistakes.

More people on Kanvis

Browse more peopleCreate your own Kanvis