// Geoinformatics Engineer

QANDEEL
FATIMA.

Maps · Models · Machines —
from raw satellite data to production ML.

01 — ORIGIN THE BEGINNING

it started as play.

BEFORE I KNEW

THE WORD

GEOINFORMATICS

I WAS DRAWING

MAPS

FOR MIDDLE-EARTH

ON COFFEE PAPER.

Geoinformatics Engineering 2022 — 2026 National University of Sciences & Technology

2022 Geoinformatics Engineering IGIS, NUST — Islamabad

My first real steps were unglamorous: ArcGIS open in front of me, spatial patterns, vector data analysis. I didn't know what I was doing. I just kept going.

The coffee stains became coordinate systems.

The drawn coastlines became datasets.

The question finally had a home.

o' ME'

it planted a question. MIDDLE-EARTH DID THIS.

Now those maps are real systems — satellite & LiDAR pipelines, ML for building footprints and crop-yield, and computer-vision work like NeRF scene-change detection.

SECTION 02 / FIELD LOG

PROJECTS

SYSTEM_ACTIVE // Short on time? Jump to Best Work ↗

GIS · RISK ANALYTICS

Black Sky Sentinel

Analyst + Developer

// SOUTHERN PUNJAB [ EXPAND ]

District Risk Intelligence

Combined real-world indicators with AHP-weighted scoring across Southern Punjab districts.
Interactive dashboard with layer toggles, per-district drill-down, and weight tuning in-browser.
Shipped as a static HTML dashboard — no backend, fully client-side.

VIEW REPO ↗

DEPLOYED

DISTRICTS

RISK LAYERS

AHP

SCORING METHOD

DEEP LEARNING · REMOTE SENSING

Building Footprint Extraction

ML Engineer

// PYTHON [ EXPAND ]

Segmentation Pipeline

End-to-end deep-learning system extracting building footprints from satellite imagery.
Automated preprocessing → training → inference loop, scalable to large regions.
Vector output ready for downstream urban-growth and census analytics.

VIEW REPO ↗

VERIFIED

PIPELINE

SAT

IMAGERY SOURCE

STACK

REMOTE SENSING · DISASTER RESPONSE

Wildfire Detection + Burn Severity

Geospatial Analyst

// FIRMS × SENTINEL-2 [ EXPAND ]

Burn-Severity Mapping

Fused NASA FIRMS active-fire feeds with Sentinel-2 pre / post imagery.
Computed dNBR and classified burn severity: unburned → low → moderate → high.
Per-fire perimeter statistics — area, severity distribution, days-since-detection.

VIEW REPO ↗

ARCHIVED

SEVERITY CLASSES

SENTINEL-2

FIRMS

NASA FEED

REMOTE SENSING · CLIMATE

Urban Heat Island · Pakistan

Geo Data Engineer

// MODIS MOD11A2 [ EXPAND ]

LST Preprocessing

Land Surface Temperature analysis using MODIS MOD11A2 8-day composites.
Pipeline handles cloud masking, QA filtering, scale-factor conversion, and per-city time-series extraction.
Outputs city-level UHI intensity trends for Pakistan's urban centres.

VIEW REPO ↗

PUBLISHED

MOD11A2

MODIS PRODUCT

8-DAY

COMPOSITE

STUDY AREA

04 — BEST WORK

ONYX

ONYX — a spatiotemporal AI framework for padel analytics. One overhead fisheye camera; full match intelligence in real time.

Homography pipeline maps fisheye pixels → metric court coordinates (sub-cm accuracy)
YOLOv11-nano ball detector at 60 fps — tracks occlusion across net, glass, players
CNN+LSTM stroke classifier: lob / smash / volley / drop / bandeja / víbora
ESP32 wrist wearable — IMU fusion streams racket kinematics over BLE
Three-tier architecture: edge (Pi 5) → local LAN server → optional cloud sync
Shot-graph rally embedding; vector DB stores 1.2 M rallies for similarity search
Player heatmaps, zone-dominance radar, serve-pattern polar charts
Sub-45 ms end-to-end latency on NVIDIA Jetson Orin NX (local inference)
Multi-match season dashboard — Progressive Web App, offline-first
Dataset: 340+ match hours labeled with court-relative bounding boxes