Atabak Mardan, PhD - Senior Data Scientist & AI Product Builder
Shipping Intelligent
Data Products at Scale.
I design and deploy end-to-end data products, from interactive dashboards to production Machine Learning, GenAI Agentic Workflows, and Spatial Analytics. Backed by a PhD in Transportation Systems, I work directly with stakeholders to turn complex ideas into shipped software, whether writing the code myself or leading cross-functional teams to deliver it.
Core Competencies
Deployed high-precision predictive models on 50M+ records. Expert in CLV optimization, time-series forecasting, and probabilistic Monte Carlo simulations.
Architecting multi-stage agentic LLM pipelines for automated intelligence, semantic vector search, and scalable feature engineering.
Designing resilient ETL architecture, data warehousing (Redshift/Glue), and massive-scale entity resolution algorithms achieving 96% accuracy.
Skills & Stack
Work Experience
Data Scientist III
Choice Hotels International
- Sales Lead Scoring Platform: Built a product that scores and ranks prospective B2B customers for hotel sales teams. Cut sales lead research from an hour to minutes for teams covering 6,000 properties. Delivered as a Tableau dashboard integrating Salesforce, internal performance data, and external vendor data. Engineered multiple ML and LLM-based features that feed a single ranked score.
- Customer Lifetime Value Model: Built an XGBoost CLV model on 50M+ transaction records. Improved customer segment accuracy by 25%. Informed targeting decisions on $100M+ in system fees for marketing and seasonal planning. Built a deduplication pipeline to clean the underlying data.
- Vision-LLM Marketing Photo Compliance Pipeline: Built an AI system that scores and ranks marketing photos across 6,000 franchised hotel properties against brand and compliance rules. A contractor quoted $45K and 6,090 hours to do this manually. Delivered the same coverage in a single automated run for under $1K, a 97% cost reduction. Flags violations including bathroom hero images, parking-lot-heavy exteriors, competitor logos, and guests without releases. Routed full-portfolio runs through Bedrock Batch Inference and downsampled images to reduce per-call token cost.
- External Data Matching Pipeline: Designed a multi-stage algorithm to match external data sources against 60,000 US hotels. Replaced a single-step name-matching approach. Added fuzzy matching, geospatial matching, and LLM-based vector similarity for hard cases. Lifted match rate from 50% to 96%. Deployed on AWS Glue.
- Climate Risk & Opportunity Dashboard: Built a first-of-its-kind tool for area directors covering 6,000 properties. Replaced ad-hoc property reviews with a single, unified view. Translated raw climate, solar, EV charging, and national park data into user-friendly KPIs (high, medium, low). Layered in GenAI (Claude through AWS Bedrock) to generate actionable, property-specific summaries.
Lead Transportation Networks Data Scientist
ICF International
- Built a Python and GIS pipeline to audit GTFS data quality across FHWA's national transit inventory. Flagged misaligned road geometry, missing stops, and other structural errors. Ranked networks by severity so the worst offenders could be addressed first. Delivered findings via technical documentation and stakeholder dashboards.
- Built Python and GIS isochrones to measure walking and biking access to bus stops along the Bronx highway corridor for NYSDOT. Mapped current transit access gaps. Modeled scenarios for proposed north/south access restrictions. Quantified how the changes would affect resident access to transit.
- Powered the backbone of Virginia OIPI's public data inventory by computing statewide travel-time reliability KPIs in R. Partnered directly with stakeholders to define KPI methodology. Delivered detailed methodology reports for client review.
Transportation Systems Data Modeler
C&M Associates
- Owned the data and simulation pipeline behind long-term traffic and revenue forecasts for state DOT projects, including the I-495 toll road extension. Ran data ingestion into the company data warehouse, outlier detection, survey processing, travel-demand modeling, and calibration in Python, GIS, and Excel.
- Built revenue forecasts and sensitivity analyses for the Dulles Toll Road 30-year planning project, feeding VDOT's long-term contracting and operator decisions. Modeled toll rate scenarios and demand sensitivities in Python and advanced Excel, working directly with VDOT stakeholders.
Data Science Graduate Research Assistant
George Mason University
- Developed a VDOT Tableau dashboard to visualize and expose intersection-level traffic metrics, enabling improved operational awareness for state transportation stakeholders.
- Parsed and cleaned raw sensor text feeds into a structured, queryable database to support downstream data analysis and reporting requirements.
- Delivered an end-to-end dashboard solution by transforming raw sensor data into actionable stakeholder insights, streamlining traffic monitoring and reporting processes.
PhD, Transportation Systems
George Mason University (2025)
MS, Data Science
George Mason University (2019)
MS, Transportation Systems
Middle East Technical University (2015)
BS, Civil Engineering
Azad University (2009)
Transportmetrica A · 2026
Case Studies on Transport Policy · 2024
Transportmetrica A · 2024
Transportation Research Part B · 2021
Personal Projects
Interactive systems, models, and analytics dashboards built to explore public data and showcase data science solutions.
Always open for collaborations!
Always ready for comments, suggestions, and collaborations on a product. Email me to pitch your idea!
Games Showcase
Interactive voxel experiences and game experiments.
Atabak & Alma Mardan
Games on this page were developed in collaboration with Product Owner Alma Mardan.
MineGURU: Retro Voxel Parkour
Codeveloped with my 7-year-old daughter Alma (Product Owner)
A fun voxel game co-developed with my 7-year-old daughter Alma (Product Owner), and fully vibe coded.
Jump Brain: Retro Runner
An 8-bit retro runner by Alma
An addictive 8-bit retro runner game by Alma. Select your champion and jump over obstacles to keep the brain jumping!
Atabak Mardan, PhD
PhD in Transportation Systems & Operations Research
I received my PhD in Transportation Systems from George Mason University, with my doctoral thesis focusing on the modeling and optimization of dynamic pricing toll roads (Express Toll Lanes). My research covers traveler route choices, traffic simulation under dynamic fees, and travel time reliability. I specialize in designing and deploying operations research frameworks, statistical learning, and spatial analysis tools.
I am highly passionate about transport economics, traffic flow physics, and network optimization, and I am always ready to collaborate, discuss new ideas, and consult on related data products.
Selected Publications & Research Output
| Publication Title | Citations | Year |
|---|---|---|
Google Scholar Profile
(3 years post-defense)
Pulled live from Google Scholar
View Live Scholar ProfileInsights & Activity
Thoughts on Applied AI, Data Engineering, and Building Products.
Get in Touch
Let's discuss predictive systems, operations research, spatial modeling, or data architecture.
Let's build the next generation of data products.
Have an idea for a project, a suggestion, or a potential collaboration? Send a message below or use the direct email link as a backup.
Message Sent!
Thank you for reaching out, . I have received your message and will get back to you within 24 hours.