Click the earth to unearth what lies beneath
Enter Portfolio →

MASOOD
SULTAN

Geoscientist & AI Engineer - Berlin · Open to PhD positions

Explore my work View Research About me
01 - Background
Masood Sultan

My career didn’t begin in tech.
It began underground.

I began as a geophysicist, learning to interpret incomplete evidence across 3D seismic volumes, geological models, and well logs. I earned a BSc in Geophysics from Bahria University with a 3.55/4.00 CGPA and placed 3rd in the AAPG Imperial Barrel Award, Asia Pacific, working in a multidisciplinary team to turn subsurface data into a defensible geological interpretation.

At Humboldt University of Berlin, my MSc in Global Change Geography expanded that systems perspective into climate adaptation, earth observation, spatial analysis, and computational modelling. My thesis examined why participatory water governance succeeds or fails under climate stress, synthesizing 121 peer-reviewed studies into a practical framework for knowledge, institutions, equity, and adaptive capacity.

That bridge between physical systems and software now defines my work. I design Python and TypeScript products that combine public APIs, browser automation, geospatial analysis, and LLM-assisted extraction. Across them, I prioritize traceable evidence, privacy, explicit uncertainty, and bounded automation over opaque claims of intelligence.

I created Arbeit.fit and lead its engineering. The live evidence-based job matching platform turns CVs, resumes, and selected public professional sources into privacy-masked profiles, prepares country-aware search plans, and ranks vacancies with exact source links so every recommendation can be checked.

Alongside my research and product work, I am the founder and lead engineer of Maven Webcraft. We build accessible, high-performance websites with hand-crafted interfaces, serverless backends, and durable SEO foundations.

Technical Stack

Languages & Frameworks

  • Python 3.10+
  • TypeScript / JavaScript
  • Node.js / Express
  • HTML5 / CSS3
  • R / RStudio

AI & Automation

  • LLM-Assisted Extraction
  • OpenAI-Compatible APIs
  • Browser Automation
  • Structured Validation
  • Notification Workflows

Data & Geospatial

  • QGIS / Spatial Analysis
  • Petrel / 3D Geomodelling
  • Earth Observation & Climate Data
  • USGS / NASA / NOAA Feeds
  • SQLite / Pandas / NumPy

Cloud, Backend & Quality

  • Appwrite Functions / Auth
  • REST APIs / SSE
  • Docker / Hugging Face Spaces
  • Pytest / Ruff / Pyright
  • Git / GitHub Actions
02 - Selected Work

Things I've Built

Research systems, not just demos

Computational geoscience tools for live hazards, adaptive scraping, intelligence fusion, and climate workflows.

Discuss research collaboration →
01
PythonAppwriteManaged AIPrivacyJob SearchArbeit.fit

Arbeit.fit

A live, privacy-conscious evidence-based job matching platform. Candidates add a CV, resume, or selected public professional pages; the system keeps source provenance separate, masks personal data before managed AI, builds a reusable profile, and searches a country-aware set of public vacancy sources. Matches are ranked with exact source links rather than opaque recommendations.

Privacy-Aware ProfilesPDF, DOCX, text, and selected public sources with PII masking before managed AI
Country-Aware SearchTen-destination plans combine authorized feeds, public career boards, and available partners
Evidence-Bound RankingExact source links and explicit uncertainty keep every match auditable
ProblemManual job search is fragmented across boards with repetitive screening and zero explainability.
RoleDesigned profile synthesis, source isolation, search orchestration, ranking, authentication, and Appwrite delivery.
StackPython, Appwrite Functions and Auth, SQLite, managed LLM integrations, HTML, CSS, and JavaScript.
ProofThe production product and public documentation are live at Arbeit.fit; the source code is private.
Visit Arbeit.fit →
View Terminal Output
02
TypeScriptNext.jsOSINTMapLibre GL

OmniWatch

A global OSINT & threat intelligence platform that fuses 48+ open data sources — military ADS-B aircraft tracking, maritime AIS vessel monitoring, USGS earthquakes, NASA wildfires, GDELT conflict events — into a single real-time dashboard with an AI correlation engine powered by local LLMs. Built on the paradigm that geoscience monitoring and intelligence analysis share the same architectural DNA.

48+ Sources Fused8-tier intelligence pipeline from hazards to SIGINT
AI Correlation EngineLLM-powered cross-referencing via Ollama
Live Map DashboardMapLibre GL with layer toggles & severity filtering
ProblemOpen intelligence and hazard signals are scattered across incompatible feeds.
RoleBuilt the fusion architecture, map layers, and AI correlation workflow.
StackTypeScript, Next.js, MapLibre GL, open feeds, and local LLM inference.
Proof48+ sources, live dashboard, and public repository demonstrate the system.
View repository → Live Dashboard →
View Terminal Output
omniwatch-server
$ npx ts-node src/index.ts
[engine] OmniWatch Intelligence Engine v1.0 booting on port 4100
[sweep] Intelligence sweep initiated — 48 sources across 8 tiers
[tier-1] USGS: 12 seismic events | NASA FIRMS: 847 fire detections
[tier-2] ADS-B: 23 military aircraft tracked | OpenSky: active
[tier-3] MarineTraffic AIS: 156 vessels in monitored zones
[ai] Correlation engine: 3 cross-source anomalies detected
[sse] Broadcasting delta to connected clients. Next sweep in 900s.
$ █
03
TypeScriptNode.jsExpressSSE

TerraMind Core

A real-time global disaster intelligence platform entirely powered by open government APIs. Dynamically aggregates data from USGS Earthquakes, NASA EONET Wildfires, NOAA Weather Alerts, and NASA FIRMS Satellite Fire Detection into a unified stream. Features a built-in GeoScience AI Assistant.

4 Gov APIs UnifiedMulti-source normalizer pipeline
Server-Sent EventsReal-time push without polling
Interactive DashboardLive map markers + severity filters
ProblemLive hazard feeds are fragmented across agencies.
RoleDesigned the API pipeline, normalizers, and dashboard flow.
StackExpress API, SSE updates, USGS, NASA, and NOAA feeds.
ProofLive dashboard and repository show the working system.
View repository → Live Dashboard →
View Terminal Output
04
PythonStructured DataBounded RecoverySQLiteMIT License

AI Scraper Prime

An audited AI-assisted extraction prototype that converts rendered pages or supplied HTML into schema-shaped data through OpenAI-compatible models. It combines typed failures, bounded recovery, a transparent heuristic quality proxy, and domain-and-schema-scoped SQLite strategy memory. Prompt refinements require measured improvement, cleaning rules are reversible, and unsupported access-control challenges escalate instead of being bypassed.

Schema-Shaped ExtractionBrowser-backed and supplied HTML paths share one tested API
Measured Strategy MemoryVersioned prompts are accepted only when the quality proxy improves
Bounded RecoveryTyped failures drive controlled retries or explicit escalation
ProblemExtraction workflows fail when layouts or providers change, while opaque quality scores hide uncertainty.
RoleBuilt the extraction runtime, strategy memory, quality proxy, recovery workflow, and CLI/API surfaces.
StackPython 3.10+, Chrome, Beautiful Soup, lxml, SQLite, OpenAI-compatible APIs, Click, and Rich.
ProofThe public v2.0 repository documents its claim matrix and offline CI gates; live-site compatibility is not implied.
View repository →
View Terminal Output
05
PythonBerlinCrawlingSQLiteNotificationsAGPL-3.0

OpenHouse Bot Prime

An audited, Berlin-focused property-listing workflow for source discovery, crawling, filtering, persistent deduplication, and configurable notifications. Source-specific crawlers feed a generic parser and optional managed fallbacks, while typed failures, scoped circuit breakers, and bounded recovery keep repeated runs controlled. It is a sequential workflow, not a claim of universal coverage or independently proven 24/7 operation.

Berlin Source Catalog139 catalogued sources; the priority scope has 86 entries and 81 crawlable URLs
Cross-Run DeduplicationSQLite-backed listing claims prevent repeated notifications across cycles
Operational GuardrailsTyped failures, scoped circuit breakers, bounded recovery, and escalation
ProblemHigh-volume listings disappear before manual search catches them.
RoleBuilt crawler orchestration, source inventory, filtering, deduplication, notifications, and operational recovery.
StackPython 3.10+, browser and HTTP crawlers, Beautiful Soup, lxml, SQLite, YAML, and configurable notifiers.
ProofThe public audited successor exposes a generated source report, implementation matrix, and offline CI gates; active monitoring is not claimed.
View repository →
View Terminal Output
Want to evaluate the work?

Start with the live products, demos, and public repositories, then reach out about PhD, research engineering, or climate AI collaborations.

Contact Masood →
03 - Web Projects

Maven Webcraft Websites

01
HTMLViteVanilla JSVanilla CSS

Café Zwei Freunde

A polished, performant single-page website for a premium coffee and bakery shop in Berlin. Features a responsive menu, modern design aesthetics, and optimized load times.

Optimized AssetsPreloaded background images and fast-loading next-gen formats
Premium Brand DesignHarmonious colors (Outfit, Playfair Display) and responsive layouts
Interactive ExperienceSmooth scrolling, custom UI components, and dynamic animations
Visit Website →
02
Next.jsTailwind CSSFramer MotionPhotoSwipe

Kindly Berlin

A premium Charlottenburg-based cafe website built with modern frontend frameworks. Features custom interactive elements, motion transitions, a dynamic menu layout, and optimized SEO.

Framer Motion AnimationsSmooth interactive micro-animations and transitions
Specialty Menu UIInteractive categorised sections for matcha, specialty coffee, and smoothies
PhotoSwipe GalleryOptimized, touch-responsive image gallery showcasing the cafe vibe
Visit Website →
03
HTMLCSSJavaScriptBilingual

Barista Portfolio

A bilingual microsite showcasing specialist barista skills and freehand latte art, featuring sequential process storytelling, a visual gallery, and an integrated ambient audio player.

Bilingual InterfaceSeamless English/German toggle for local market reach
Latte Art GalleryCurated visual showcases of precise freehand latte art
Audio ExperienceIntegrated custom audio engine playing ambient cafe tracks
Visit Website →
04
Next.jsTypeScriptReactTailwind CSS

Shareekat-ul-Hussain

A bilingual Urdu-first platform for Masjid-o-Imambargah Shareekat-ul-Hussain in Karachi, combining a Quran reader, date-aware Karachi prayer times, Shia calendar events, programs, and location flows.

Quran ReaderLazy-loaded surahs with Uthmani Arabic text and Urdu translation
Shia CalendarInteractive monthly calendar with celebration/mourning theme overlays
Bilingual & RTLSeamless RTL Urdu-first presentation and English translation routes
Visit Website →
05
Next.jsTypeScriptTailwind CSSMarkdown

Masood Sultan Blog

An editorial blog connecting AI systems, computational geoscience, climate modelling, and open-source intelligence through clear research trails and long-form technical writing.

Editorial IdentityClean, focused typography for research trails and long-form essays
Markdown DrivenStatic rendering of complex geoscientific technical content
Responsive LayoutSeamless reading experience optimized across all devices
Visit Website →
04 - Publications

Academic Research

Master’s Thesis 2025 Humboldt University of Berlin Global Change Geography

A Critical Review of Participatory Approaches in Water Management for Climate Change Adaptation

Supervisors: Prof. Dr. Tobias Krüger & Prof. Dr. Dieter Gerten · Geographisches Institut, Humboldt-Universität zu Berlin · 102 pages

Climate change is dismantling long-standing assumptions of stable hydrological patterns, exposing institutional fragmentation and amplifying inequality in water governance. This thesis develops the Conditional Enabling Framework (KIEA) - a diagnostic instrument that identifies when participation leads to effective, equitable, and adaptive water governance under climate stress.

Through a realist synthesis of 121 peer-reviewed documents (2009-2025), the research demonstrates that participation often devolves into tokenism or elite capture unless four interdependent enabling conditions are met simultaneously:

K - Knowledge PerformanceData availability, scientific literacy, and knowledge co-production among stakeholders
I - Institutional PerformanceGovernance capacity, regulatory design, and formal/informal institutional coherence
E - Equity PerformanceEquitable stakeholder representation, inclusive decision-making, and distributive justice
A - Adaptive PerformanceFlexibility, iterative learning, and system resilience under uncertainty

When any single dimension falls below threshold - the binding constraint - the entire governance system underperforms, regardless of strength in other areas. This weakest-link logic directly shifts theoretical climate governance toward measurable, actionable, and prescriptive implementation.

Download Full Thesis (PDF) →
Journal of Applied Geophysics 2020 Elsevier 3D Geomodelling

A Case Study of 3D Geomodelling of Frontier Formation Second Wall Creek Sand, Teapot Dome, Wyoming, USA

Co-author · Khan, H.A., Sultan, M., Khan, M.J., Alvarez, M.D., Mehdi, S.D. & Javed, M.A.

An advanced 3D geomodelling workflow applied to the Teapot Dome anticline within the Powder River Basin - one of the largest producing basins in the United States, with 2.3 TCF gas and over 2.7 billion barrels of recoverable oil. Using integrated well log correlation, seismic interpretation, and stochastic property modelling in Schlumberger Petrel, the research demonstrates best practices in reservoir characterization for the Second Wall Creek Sand of the Frontier Formation.

Sultan, M. et al. (2020). Journal of Applied Geophysics, Vol. 179, 104114.

DOI: 10.1016/j.jappgeo.2020.104114 → Download Manuscript (PDF) →
Bachelor’s Thesis 2018 Bahria University Petroleum Geoscience

3D Seismic Geomodelling of Cretaceous Shoreface Reservoir Sands, Frontier Formation, Teapot Dome, Wyoming, USA

BSc Geophysics · Bahria University, Karachi · 74 pages

The foundational geoscience research that seeded the Elsevier publication. This thesis focuses on the 3D seismic geomodelling of Cretaceous shoreface reservoir sands within the Teapot Dome — a publicly available dataset widely used in petroleum geoscience education. The work integrates seismic volume interpretation, well log data, stratigraphic correlation, and property modelling to build a comprehensive subsurface reservoir model.

Download Full Thesis (PDF) →
04 — Downloads

Media & Resources

Documents

📄
Academic CV Masood Sultan — Geoscientist & AI Engineer · Berlin
Download PDF →

Academic Publications

📘
MSc Thesis — A Critical Review of Participatory Approaches in Water Management for Climate Change Adaptation Humboldt University of Berlin · Global Change Geography · 2025 · 102 pages
Download PDF → 📗
Research Paper — A Case Study of 3D Geomodelling of Frontier Formation, Teapot Dome, Wyoming Journal of Applied Geophysics · Elsevier · 2020 · 31 pages
Download PDF → 📙
BSc Thesis — 3D Seismic Geomodelling of Cretaceous Shoreface Reservoir Sands, Teapot Dome Bahria University · Geophysics · 2018 · 74 pages
Download PDF →

Looking for a PhD position
in computational geoscience
or climate AI.

If you're seeking a researcher at the intersection of geoscience, AI, and climate adaptation - someone who can build the tools and do the science - let's connect. Based in Berlin. For the fastest reply, include the role, lab, project, timeline, and links to any relevant call or research group.

Open to research positions
Download Academic CV →
Email about research roles Review CV and papers
Email masood.geo[at]yahoo.com → LinkedIn /in/masood1996 → GitHub masood1996-geo →