Joseph McDevitt | Finance & AI

About Me

Finance professional building AI-powered tools for deal sourcing, acquisitions, and private markets. University of Miami, cum laude.

Projects

Live Platform

Pharmacy M&A Intelligence Platform

A full-stack acquisition targeting platform that ingests 112,000+ U.S. pharmacy records from the CMS NPI registry, enriches with Medicare Part D claims and Census demographics, classifies chain vs. independent, scores acquisition readiness using a 6-factor weighted model, and tracks deals through an 8-stage pipeline.

Interactive Demo

Problem

Pharmacy acquisition teams manually sift through fragmented public databases to find independent targets. There's no centralized tool that combines provider registry data, prescription volume, Medicare/340B designation, demographics, and competitive density into a single scoring system — and no way to probabilistically identify which stores are most likely to be available at a discount.

Solution

An automated pipeline that downloads the full NPPES registry (~700MB, 10GB+ expanded), filters across 12 pharmacy taxonomy codes, classifies 112,000+ pharmacies as chain vs. independent, enriches with Medicare Part D claims and ZIP-level Census demographics, applies a weighted acquisition score, surfaces closing signals and retirement risk probabilities, and delivers everything in a filterable dashboard with a built-in deal tracker.

How It Works

NPI Registry Ingestion

Downloads and processes the full CMS NPPES registry — 112,000+ pharmacies across all 50 states and territories

Source: CMS NPPES Data Dissemination — downloaded via httpx streaming in 1MB chunks with a 1-hour timeout
Parsing: Pandas read_csv with chunksize=100,000 rows and low_memory=False. Entity type filter (organizations only). Batch inserts of 5,000 records per database commit
12 pharmacy taxonomy codes: Community/Retail, Compounding, Long Term Care, Home Infusion, Institutional, Mail Order, Military, Managed Care, Nuclear, Specialty, General Pharmacy, and Pharmacist
Normalization: Phone standardized to (XXX) XXX-XXXX, names uppercased, ZIP truncated to 5 digits, dedup key = MD5 hash of org_name|address|zip
NPI date extraction: Enumeration date, last update date, deactivation date/reason parsed from columns 36–39. Years in operation calculated from enumeration date
Coverage: 64 states/territories. Top 5 independent: NY (6,827), CA (5,844), TX (5,502), FL (5,248), MI (2,819)

Chain Classification Engine

Identifies and separates chains, independents, and institutional pharmacies using pattern matching and multi-location clustering

Named chains matched: CVS, Walgreens, Walmart, Rite Aid, Kroger, Costco, Sam's Club, Target, Publix, H-E-B, Albertsons, Safeway, Meijer, Winn-Dixie, Giant, ShopRite, Wegmans, Hy-Vee, Fred Meyer, Harris Teeter, Omnicare, PharMerica, Kindred, BrightSpring, Cardinal Health, McKesson, AmerisourceBergen, Express Scripts, Optum Rx, Cigna, Amazon Pharmacy, Capsule, Alto, Genoa, PharmHouse
Multi-location clustering: Any organization name appearing at 10+ distinct addresses is auto-flagged as a chain — catches regional operators not in the named chain list (~18,000 pharmacies)
Institutional filtering: Regex patterns for: Hospital, Medical Center, Nursing, Long-Term Care, LTC, Skilled Nursing, Rehab, Assisted Living, Infusion, Correctional, Prison, Veterans, VA — tagged but not deleted from database
Ownership signals: Corporation (41,764), LLC (30,449), Unknown (39,174), Partnership (935), Professional Corporation (215)
Result: 65,279 independents (58%), 47,258 chains (42%), 3,890 institutional

Medicare & 340B Enrichment

Enriches pharmacies with Medicare Part D claims data and identifies 340B-eligible pharmacies — key signals for file value and acquisition pricing

Medicare Part D fields: Total claims count (prescriptions filled), unique Medicare beneficiaries served, total drug cost in dollars — sourced from CMS prescriber-level utilization data
340B program: Federal drug pricing program that requires drug manufacturers to sell outpatient drugs at a discount to eligible healthcare organizations. 340B pharmacies acquire drugs at 25–50% below retail — dramatically affecting per-script margin and file value calculations
340B eligibility signals: Federally Qualified Health Centers (FQHCs), critical access hospitals, disproportionate share hospitals, contract pharmacy arrangements — identified via taxonomy codes and organization name patterns
Why it matters for M&A: A pharmacy's 340B status signals margin compression risk for buyers — or opportunity if the acquirer can maintain the 340B contract. Medicare volume indicates patient stickiness and reimbursement predictability

Census Demographic Scoring

Enriches every pharmacy with ZIP-level Census data — aging population, income, growth, and competition density — to identify high-value markets

Coverage: 99.8% of pharmacies enriched with Census data (112,154 of 112,537)
% Age 65+: Range 0–100%, average 17.93%. Scaled to a 30% threshold (30% or higher = max score of 100). Rationale: seniors fill 2–3x more prescriptions and have higher adherence — directly correlates with script volume and file value
Median household income: Range $2,499–$250,001, average $81,776. Scaled to $100K threshold. Higher income = better commercial payer mix = higher reimbursement per script
Population growth: Weighted multiplier of 5x applied (e.g., 10% growth → 100 pts). Signals expanding patient base and market demand
Competition density: Pharmacies per 10,000 residents per ZIP. Range 0.17–10,000, average 6.65. Thresholds: ≤1 → 100 pts (excellent), ≤3 → 80, ≤5 → 60, ≤8 → 40, ≤12 → 20, >12 → 10 (saturated)
File value estimate: Rx volume × $4/script (industry standard acquisition price). Total independent pipeline value: ~$4.72 billion

6-Factor Acquisition Score

Composite 0–100 score combining Rx volume, competition, aging population, retirement risk, income, and population growth

Rx Volume (30%): Estimated scripts/year scaled to 80,000 baseline = 100 pts. Range across database: 1,732–123,750 scripts/year, average 18,073. Default 20 pts if volume unknown
Competition (20%): Based on zip_pharmacies_per_10k. ≤1 → 100, ≤3 → 80, ≤5 → 60, ≤8 → 40, ≤12 → 20, >12 → 10
Aging Population (20%): % residents 65+ scaled to 30% ceiling. Average score strong due to 17.93% national average
Retirement Risk (15%): Based on years since NPI enumeration. ≥25 yrs → 100 (owner likely 60+), ≥20 → 80, ≥15 → 50, ≥10 → 30, <10 → 10. 3,435 pharmacies currently at 20+ years
Income (8%): Median ZIP income / $100,000 × 100. Caps at 100
Pop Growth (7%): 50 + (growth_pct × 5). 0% growth = 50 pts, 10% growth = 100 pts
Final score distribution: Range 9.69–87.48, average 44.41 across all 112,537 pharmacies

Closing Signals & Probability

Surfaces pharmacies with a high likelihood of being available or discounted based on retirement risk, record staleness, and nearby closures

Long-tenured (20+ years): 3,435 pharmacies. Owner likely age 60+ and approaching retirement — highest probability of willingness to sell. NPI enumeration date used as proxy for business age
Stale NPI records (3+ years): 71,986 pharmacies. No update to CMS registration in 3+ years may indicate declining engagement, reduced operations, or owner disengagement — early signal of potential availability
Nearby deactivations: 155 pharmacies recently deactivated in the system. Platform surfaces independent pharmacies in the same ZIP code — these stores inherit displaced patient volume, making them higher-value targets (and their owners may be motivated by the same market pressures)
Change detection: Snapshots full database state before each pipeline run. Compares new vs. previous records to detect new registrations, field updates, and deactivations. Stores audit trail with change type, field name, old value, and new value
Pipeline tracking: Each run logs started_at, completed_at, status, records_processed, records_added, records_updated, and changes_detected

Key Highlights

112,000+ pharmacies ingested, classified, and scored across all 50 states
Multi-source enrichment: CMS NPI registry, Medicare Part D claims, 340B designation, U.S. Census demographics
6-factor weighted acquisition score with retirement risk probability and closing signal detection
Change detection system snapshots state before each run and tracks field-level deltas
8-stage deal pipeline tracker (Prospect to Closed) with contact management and CSV export
Dockerized full-stack deployment — FastAPI, React, PostgreSQL in one command

Technologies

Python FastAPI React PostgreSQL Docker SQLAlchemy Tailwind CSS Leaflet Maps Pandas CMS / NPI Data Medicare Part D U.S. Census API

Live Platform

Deal Sourcing Agent

An AI-powered deal origination engine that spans three verticals — PE acquisition targets, venture-backed startups, and PI law firm acquisitions. Aggregates deal flow from 20+ data sources, scores opportunities using multi-factor algorithms, enriches with Census/BLS/CMS/ARDC public data, and surfaces ranked pipelines across interactive dashboards deployed on Vercel.

Interactive Demo

Problem

Deal sourcing across PE, venture, and niche verticals is fragmented and manual. Analysts scan dozens of platforms, apply inconsistent criteria, and miss time-sensitive opportunities. There's no unified system to aggregate, score, and rank deal flow across multiple asset classes using real public data.

Solution

A multi-vertical deal origination platform with three specialized engines — PE acquisition targets (8 marketplace sources, dual scoring, Census/BLS enrichment), venture startup detection (19 data sources, 6-dimension scoring, 82 VC firm matching), and PI law firm acquisitions (ARDC retirement radar, off-market deal flow) — all surfaced through interactive dashboards on Vercel.

How It Works

PE Deal Engine

Aggregates 2,300+ deals from 8 broker marketplaces across 35 states with dual scoring for operator vs. PE buyers

8 marketplace sources monitored daily — extracts asking price, EBITDA, revenue, employee count, industry, location
Dual scoring: Independent 0-100 frameworks for solo/owner-operator buyers vs. PE sponsors (50 pts qualitative + 50 pts quantitative each)
Census Bureau + BLS enrichment: Industry survival rates, employment growth, wage pressure, competition density, market fragmentation at the county level
Interactive Leaflet map with 40+ markets, analytics dashboard, SBA financing calculator, and 2-page PDF deal reports
12 hard-exclude categories and EBITDA minimums filter out low-quality deal types automatically

Venture Intelligence

Signal-based startup detection monitoring 19 data sources, scoring companies 0-100, and matching to 82 VC firms

19 data sources monitored for startup signals — funding announcements, hiring patterns, product launches, patent filings
6-dimension scoring: Team, market, traction, product, timing, and defensibility — each weighted and scored 0-100
VC matching: Automatically matches startups to 82 VC firms across Chicago and New York based on sector, stage, and check size
CRM pipeline: 5-stage tracking from discovery through meeting, offer, and close

PI Firm Acquisitions

ARDC retirement radar identifying PI law firms for sale in Illinois through regulatory filings and off-market signals

ARDC scraping: Monitors Illinois Attorney Registration and Disciplinary Commission for retirement filings and status changes
Age-out detection: Identifies PI firms where senior partners are approaching retirement — signals acquisition opportunity
Off-market signals: Regulatory data patterns that indicate potential willingness to sell before public listing
Next.js dashboard on Vercel with firm profiles, retirement timeline, and deal pipeline tracker backed by Supabase

Key Highlights

3 verticals — PE acquisitions, venture startups, and PI law firm deals in one platform
20+ data sources aggregated with automated daily pipeline runs
Multi-factor scoring algorithms tailored to each deal type
Public data enrichment: Census Bureau, BLS, CMS, and ARDC
Live dashboards on Vercel with maps, analytics, PDF exports, and pipeline tracking

Technologies

Python Next.js TypeScript SQLite Supabase Vercel Leaflet.js Chart.js Census API BLS API NLP Engine

Active Development

Bearing x Tax

A merged SMB consulting dashboard combining a full business health analysis platform with 226 tax-saving strategies. Connects to QuickBooks for auto-fill, profiles each business across financial health, operational efficiency, and growth readiness — then matches them to applicable tax strategies ranked by estimated savings.

Interactive Demo

Problem

Small business owners don't know which tax strategies apply to them, and consultants waste hours profiling each client manually. There's no system that connects live financial data to a curated knowledge base of tax-saving opportunities and automatically matches the right strategies to each business.

Solution

An integrated platform that pulls real financial data from QuickBooks, runs a multi-dimensional business health analysis (financial, operational, growth), and matches the business profile against 226 tax strategies — each backed by IRC sections, Treasury Regulations, and court opinions — surfacing the highest-value opportunities first.

How It Works

Business Health Analysis

Multi-dimensional profiling across financial health, operational efficiency, and growth readiness

Financial health: Revenue trends, margins, cash flow, debt ratios, and working capital analysis
Operational efficiency: Employee productivity, overhead ratios, and process bottlenecks
Growth readiness: Scalability indicators, market position, and expansion potential
QuickBooks auto-fill: Pulls live P&L, balance sheet, and cash flow data directly from the client's books

226 Tax Strategy Engine

Curated knowledge base of tax-saving strategies backed by primary legal authority with eligibility matching

226 strategies: Each backed by IRC sections, Treasury Regulations, and relevant court opinions
Eligibility logic: Fact-dependency decision trees determine which strategies apply to each business
Savings estimation: Computation models estimate dollar-value savings for each applicable strategy
Priority ranking: Strategies sorted by estimated savings — highest-impact opportunities surfaced first

Integrated Dashboard

Unified consulting interface combining business diagnostics with actionable tax strategy recommendations

Client profiles: Complete business snapshots with health scores and key metrics
Strategy matching: Auto-matched tax strategies with eligibility explanations and estimated savings
Consultant workflow: End-to-end flow from client intake to strategy delivery
QuickBooks OAuth: Direct integration for automated financial data ingestion

Key Highlights

226 tax strategies backed by IRC sections, Treasury Regs, and court opinions
QuickBooks integration for auto-fill from live financial data
Multi-dimensional business health scoring across 3 domains
Fact-dependency decision trees for eligibility matching
Estimated dollar-value savings per strategy, ranked by impact

Technologies

Python Streamlit QuickBooks API SQLite Tax Knowledge Base Decision Trees

Active Development

FundLens

A SaaS portfolio monitoring platform for small and mid-size family offices. Replaces manual spreadsheet tracking and $65K+/year enterprise tools with AI-powered PDF report parsing, full fund analytics, vintage analysis, J-curve modeling, and an IC deck builder — all for a fraction of the cost.

Interactive Demo

Problem

Small and mid-size family offices ($50-500M AUM) are stuck between expensive enterprise platforms at $65K+/year and error-prone spreadsheets. There's no affordable tool that automates the ingestion of quarterly fund reports, tracks performance across a multi-fund portfolio, and produces the analytics needed for investment committee meetings.

Solution

A full-stack SaaS platform with AI-powered PDF parsing that automatically extracts fund data from quarterly reports, eliminating manual data entry. Provides 13 analytics pages covering cash flows, J-curve modeling, vintage analysis, fee tracking, commitment pacing, period comparisons, and a custom chart builder — all for $1-3K/month.

How It Works

AI PDF Parsing

Upload quarterly fund reports and AI extracts NAV, cash flows, IRR, multiples, and fee data automatically

Qwen 3.5 AI: Local LLM extracts structured data from unstructured PDF fund reports — no API costs
Confirmation screen: Parsed data shown for review before committing to database
Multi-format support: Handles various GP report formats, quarterly statements, and capital call notices
Historical import: Bulk upload past reports to build complete fund histories

13-Page Analytics Suite

Full portfolio analytics: cash flows, J-curve, vintage analysis, fee tracking, commitment pacing, and custom chart builder

Portfolio dashboard: AUM overview, fund allocation, performance summary, and recent activity
Cash flow analysis: Contributions, distributions, and net cash flow over time with forecast modeling
J-curve modeling: Visualize the PE J-curve across fund vintages with expected recovery timelines
Vintage analysis: Compare fund performance by vintage year with PME benchmarking
Fee tracking: Management fees, carried interest, and total cost analysis per fund
Analytics workbench: Custom chart builder with save-to-localStorage — build any visualization on the fly

Full-Stack Architecture

Next.js frontend on Vercel, Python FastAPI backend on Railway, Supabase for database/auth/storage

Frontend: Next.js 16 with React 19 — deployed on Vercel with server-side rendering
Backend: Python FastAPI with 7 API endpoints — deployed on Railway
Database: Supabase (Postgres) with row-level security and real-time subscriptions
AI processing: Qwen 3.5 9B running locally — zero API costs for PDF parsing
Target pricing: $1-3K/month vs. Addepar's $65K+/year — 95% cost reduction for family offices

Key Highlights

AI-powered PDF parsing eliminates manual data entry from quarterly fund reports
13 analytics pages: cash flows, J-curve, vintage analysis, fees, commitment pacing, custom charts
95% cost reduction vs. enterprise tools ($1-3K/month vs. $65K+/year)
Full-stack: Next.js + FastAPI + Supabase with local AI processing
Built for $50-500M AUM family offices with multi-fund portfolios

Technologies

Next.js 16 React 19 Python FastAPI Supabase Vercel Railway Qwen AI

Live Platform

Nomad Travel Guide

An interactive travel intelligence platform covering 100 countries, 553 cities, and 8,500+ curated activities. Features an interactive world map, a proprietary Danger-O-Meter rating system, budget rankings, trip builders, food discovery, and a multi-country planner — all enriched by a dual-AI pipeline using Qwen 3.5 and Claude Opus.

Interactive Demo

Problem

Mainstream travel guides are sanitized, generic, and surface-level. There's no single platform that gives you brutally honest safety ratings, off-the-beaten-path activities, real budget data, and local food culture for 100+ countries — all in one place with an interactive map you can actually explore.

Solution

A map-first travel intelligence app that covers 100 countries and 553 cities with 8,500+ curated activities. Features a proprietary Danger-O-Meter rating system scoring safety across 5 categories, 10 global discovery tools, per-city food guides, multi-day trip builders, and a dual-AI enrichment pipeline that generates high-quality content at zero API cost.

How It Works

Interactive World Map

Explore 100 countries on a dark-themed interactive globe with color-coded regions and activity-sized markers

Map engine: PyDeck with CARTO dark basemap — no API keys or Mapbox tokens required
Visual encoding: Dots color-coded by world region, sized by activity count per city, with glowing halo effects
Navigation: Click any dot to drill into Country → City detail views with session-state routing
Coverage: 100 countries across all inhabited continents — from Japan and Iceland to Burkina Faso and Uzbekistan

Danger-O-Meter Rating System

Proprietary safety scoring across 5 categories with 10 brutally honest danger tiers

5 scoring categories: LGBTQ+ Rights, Drug Tolerance, Women's Rights, Child Labor, Religious Freedom — each rated 1-5, total score out of 25
10 danger tiers: From safest to most dangerous — a no-holds-barred ranking system with memorable tier names
Bonus stats: Foreign aid received per year and average lifespan displayed as stat bars for additional context
Why it matters: Standard travel sites hide uncomfortable truths. This system gives travelers the unfiltered picture so they can make informed decisions

10 Global Discovery Tools

Danger Index, Budget Ranker, Country Compare, Global Search, Food Discovery, Surprise Me, When to Go, Region Explorer, Stats Dashboard, Multi-Country Planner

Budget Ranker: Compare daily costs across all 100 countries — find where your dollar goes furthest
Food Discovery: Browse local cuisine across 553 cities with 7+ dishes per city
Trip Builder: Build multi-day itineraries per city with activity sequencing
Multi-Country Planner: Chain together countries into regional travel routes with time and budget estimates
Surprise Me: Random destination generator for the spontaneous traveler

Dual-AI Enrichment Pipeline

Qwen 3.5 generates initial content locally, Claude Opus re-enriches for quality — zero API cost

Qwen 3.5 9B: Runs locally on a dedicated GPU machine — generates initial country/city/activity data in chunked prompts
Claude Opus agents: Re-enriches major countries to full quality — 15 activities per city, 7 food items per city
Resume-safe: Progress tracked in JSON — pipeline can be stopped and restarted without data loss
Result: 553 cities, 8,500+ activities, all generated and curated without a single API dollar spent

Key Highlights

100 countries, 553 cities, 8,500+ activities — one of the most comprehensive travel datasets built from scratch
Proprietary Danger-O-Meter with 5-category safety scoring and 10 danger tiers
10 global tools: budget ranker, food discovery, trip builder, multi-country planner, and more
6-tab country pages: Overview, Cities & Activities, Food, Routes, Trip Builder, Practical Info
Dual-AI pipeline (Qwen + Opus) generates all content at zero API cost

Technologies

Python Streamlit PyDeck Qwen 3.5 Claude Opus CARTO Maps Streamlit Cloud

Experience

Financial Rotational Analyst

Walgreens

Chicago, IL Jan 2026 – Present

Designed and built an executive-facing analytics dashboard that consolidated audit risk data, coverage gaps, and resource allocation into a single decision-support tool for senior leadership
Integrated automated data pipelines to replace manual audit tracking processes, reducing reporting cycle time and improving the accuracy of risk scoring across business units
Applied AI-assisted analysis to surface audit priority shifts based on emerging risk patterns, enabling the audit committee to dynamically reallocate resources rather than follow a static annual plan
Worked directly with C-suite stakeholders to translate complex audit data into clear, actionable visuals — turning what was previously a quarterly static report into a living tool that leadership actively uses for planning

Internal Audit Dashboard Development AI Analytics Executive Reporting Process Automation

Analyst

Transal Corp

Miami, FL Aug 2025 – Dec 2025

Reviewed 130+ active private equity investments and screened hundreds of new opportunities within a multi-billion-dollar portfolio
Modeled capital call pacing vs. distribution flows in Excel to enhance liquidity forecasting and cash pool allocation
Utilized Addepar and advanced Excel modeling for portfolio oversight, co-investment evaluation, and risk management
Built dynamic IRR and MOIC models simulating long-term index exposure; benchmarked to internal returns

Private Equity Portfolio Analysis Financial Modeling

Specialty Pharmacy Finance Intern

Walgreens

Chicago, IL Jun 2025 – Aug 2025

Built a Power BI dashboard from millions of quarterly data points to visualize per-script drug profitability
Delivered ad hoc drug-level financial insights to therapy directors and field leadership for strategic decisions
Developed a dietary supplement vending machine concept using DCF modeling and private credit financing

Power BI Data Analytics DCF Modeling

Equity Analyst Intern

Maredin Wealth Advisors

Miami, FL May 2024 – Oct 2024

Researched equities to support an independent RIA and contributed to enhanced portfolio returns
Developed a full investment rationale for Snowflake Inc. covering business model, financials, and market potential

Equity Research Investment Analysis

Algorithmic Trader

Core Value Capital LLC

Chicago, IL Jul 2023 – Feb 2024

Conducted research and backtesting to identify new currency pairs for a proprietary trading algorithm
Applied calculus-based mathematical principles to enhance algorithm effectiveness across 13 currency pairs
Optimized capital allocation and leverage thresholds, removing 15% of pairs to mitigate excess risk

Algorithmic Trading Quantitative Analysis Risk Management

Skills & Credentials

Financial Modeling

LBO Modeling

DCF Analysis

M&A Modeling

Comparable Analysis

Precedent Transactions

Three Statement Models

IRR / MOIC Analysis

Capital Call Pacing

Mathematical Modeling

Technical Skills

Microsoft Excel (Expert)

VBA Macro Development

Power BI

SQL

Python

Addepar

Bloomberg Terminal

AI & Automation

ChatGPT / GPT-4

Claude AI

OpenAI API

GitHub Copilot

Microsoft Copilot

Cursor AI

AI-Assisted Financial Analysis

Prompt Engineering

LLM Workflow Automation

AI Data Extraction & NLP

RAG Pipelines

AI-Powered Dashboard Development

Education

University of Miami

Herbert Business School

B.S. Finance & Mathematics

Cum Laude

Provost's Honor Roll | Dean's List

Certifications

WSP

Wall Street Prep

Financial Modeling Certification

LBO, DCF, M&A, Comps, Precedents, Three Statement Modeling, Excel VBA

What Drives Me

Artificial Intelligence

I believe AI will fundamentally reshape finance — from deal sourcing and due diligence to portfolio construction and risk management. I'm actively building tools and workflows that put AI at the center of the investment process.

Buying Companies

There's nothing more exciting than finding an undervalued business, structuring a deal, and creating value through operational improvement and strategic growth. M&A and acquisitions are where I want to build my career.

Private Markets

Public markets get all the headlines, but the real alpha is in private markets. From PE and VC to private credit and co-investments, I'm drawn to the complexity and outsized returns of illiquid strategies.

Let's Connect

I'm always interested in discussing deals, AI applications in finance, or new opportunities.

Josephd.mcdevitt@gmail.com LinkedIn 847-504-7993