Pittsburgh Artificial Intelligence Investment Ecosystem – Technical Documentation

Block Center for Technology & Society, Carnegie Mellon University

Principal Investigators: Pablo Zavala · Liufei Chen
Release: v2.0 (July 2025)

---

1 Project Purpose

This repository provides a reproducible, data-driven assessment of venture-backed artificial-intelligence companies headquartered in Pittsburgh (and peer regions). The codebase ingests Crunchbase-derived datasets, rigorously cleans and enriches them, and outputs an interactive, publication-ready HTML report suitable for academic and policy analysis.

Public scope. The findings published in this repository cover the Pittsburgh metropolitan area only: 133 active AI firms and USD 6.3 B cumulative funding (July 2025 snapshot), computed by the committed pipeline from Crunchbase-derived datasets. In compliance with Crunchbase's Terms of Service, the raw exports are not redistributed in this repository (see Section 4); the committed HTML report is the pipeline's verbatim output. Peer-region data supported the report's peer-city benchmarking section only; no aggregate findings beyond Pittsburgh are published here.

2 Repository Structure

Path	Description
comprehensive_pittsburgh_analysis.py	Core ETL and metric-calculation engine (Pandas + NumPy).
unified_pittsburgh_report.py	Orchestrates visual analytics (Plotly + Folium) and compiles the single HTML deliverable.
Companies-With-Address - Sheet1.csv, *.geojson	Included geo files (hand-geocoded company points; county ZIP polygons). Raw Crunchbase exports are excluded -- see Section 4.
Pittsburgh_AI_Preliminary_Report.html	Most recent build of the interactive report (open in any modern browser).

3 Quick-start (local execution)

1. Create/activate a Python 3.10+ environment.
2. Install requirements:

pip install -r requirements.txt

3. Run the pipeline:

python - <<'PY'
from comprehensive_pittsburgh_analysis import run_comprehensive_analysis
from unified_pittsburgh_report import UnifiedPittsburghReport
an = run_comprehensive_analysis()
UnifiedPittsburghReport(an).generate_unified_report()
PY

The script regenerates Pittsburgh_AI_Preliminary_Report.html in the project root.

Note. Re-running the pipeline end-to-end requires the five raw Crunchbase export files listed in Section 4, which are not redistributed here. The committed Pittsburgh_AI_Preliminary_Report.html is the pipeline's verbatim output; the raw July-2025 snapshot is available on request to the authors for academic verification.

4 Data Inventory and Availability

File	Rows	Key variables	Availability
all-minus-austin-companies-7-18-2025.csv	1 324	funding $, industry tags, HQ location, founding dates.	excluded -- Crunchbase license
austin-companies-7-18-2025.csv	635	peer-city company universe.	excluded -- Crunchbase license
Companies-With-Address - Sheet1.csv	69	point geocodes (lat/lon).	included
funding_transactions_full.csv	5 564	round-level amounts, dates, participant counts.	excluded -- Crunchbase license
investor_profiles_unique (1).csv	5 633	investor meta-data (location, stage focus).	excluded -- Crunchbase license
pittsburgh-schools-7-23-2025.csv	68	university descriptors (alumni, founders).	excluded -- Crunchbase license
Allegheny_County_Zip_Code_Boundaries.geojson	110 polygons	ZIP polygons for choropleths.	included (public county GIS)

Source: Crunchbase (crunchbase.com), July 2025 snapshot. Raw exports are not redistributed in compliance with Crunchbase's Terms of Service; they are available on request to the authors for academic verification. All processing runs locally; the pipeline makes no external API calls.

5 Analytical Pipeline (summary)

1. Entity normalisation – deduplicate organisations, coerce monetary strings to floats, harmonise ZIP-codes.
2. Industry classification – multi-label keyword matching; proportional funding weighting across eight macro-sectors.
3. Metric derivation – company age, funding velocity, innovation score, employment estimates, unicorn flag, etc.
4. Network extraction – explode Top 5 Investors strings, construct investor–company bipartite graph; compute portfolio statistics.
5. Visual synthesis – Plotly dashboards (industry, funding stages, investor performance, peer-city benchmarking) and Folium geospatial layers (choropleths, marker clusters, university overlay).

6 Key Findings (2025-07 snapshot)

• 133 active AI firms headquartered in Pittsburgh, USD 6.3 B cumulative funding.
• Sectoral funding leadership: Autonomous Systems (38 %), Core AI/ML (34 %), Healthcare AI (7 %).
• Median time-to-first funding: 2.4 years; median employee count: 18.
• Investor landscape dominated by a small set of specialised funds; portfolio success-rate median = 66 %.
• Zip-code level clustering around 15213 (CMU/Pitt) exhibiting > USD 2 B raised.

Full methodological notes appear in the generated report.

7 Reproducibility & Contribution

The codebase is modular and annotation-rich. Researchers wishing to extend the study (e.g., additional peer regions, longitudinal updates) should fork the repo and observe the following guidelines:

1. Submit pull-requests with atomic commits and descriptive messages.
2. Adhere to PEP-8 and maintain type consistency (use mypy stubs where practical).
3. Update unit tests (pytest) if you alter core computation functions.
4. Document new variables/datasets in this README.

For questions, contact pzavalar@andrew.cmu.edu or liufeic@cmu.edu.

---

© 2025 Carnegie Mellon University – Released for non-commercial research use.