Machine Learning: Theory and Hands-on Practice with Python
University of Colorado Boulder
VPH8YNY2240CVerify
Philadelphia, PA
Felipe Molino
AI, Product, Engineering & Supply Chain Technology Executive
I turn analytics, optimization, and AI into products and services that run in production, earn their keep commercially, and stay trustworthy.
🏆 2026 FreightWaves AI Excellence in Supply Chain Award — Operational AI Integration
$200M+
in total supply chain savings delivered at NFI
$0 → 7-fig
consulting and digital twin practice built from scratch
~$700M
in customer sales my engagements supported
3 mo → 2 wks
new-account onboarding, through third-party platform integrations
260+
hours per week returned to operations by shipped AI
39
certifications in machine learning, AI ethics, generative AI and wine
About
Twelve years of turning ambiguous mandates into measurable outcomes.
I'm an AI, product, and engineering executive with 12+ years in logistics and supply chain technology. At NFI Industries I've turned analytics and optimization into revenue: I built the consulting and digital twin practice from $0 to seven figures, led 50+ network-design engagements at 10–15% average savings, and supported ~$700M in customer sales. In total, the work delivered $200M+ in supply chain savings at NFI. Over my career I've modeled more than $1B in savings, including two engagements that each delivered $100M+.
On the product side, I've owned NFI's Transportation Digital Twin since 2018 and took it from a $500K pilot to a production B2B SaaS decision platform modeling about $10M in annual savings across 10 enterprise accounts. I manage an eight-figure vendor and partner portfolio alongside NFI's President of Transportation Management, VP of Innovation, and VP of Procurement, and I cut RFP and continuous-improvement turnaround from six weeks to one.
On the engineering side, I built an internal platform and infrastructure team from 0 to 20+ across six countries, and was part of the leadership team for a 50+ engineer offshore staff-augmentation engagement (about $1M a year), where many of the engineers reported to me directly.
The work I'm proudest of is the unglamorous kind. I led a six-month phased production migration off an expiring platform license — running live and validation workloads in parallel, then cutting over account by account across ten enterprise accounts with no service interruption. I ship AI into live transactional systems rather than demos: an XGBoost tender-acceptance model writing to Navitrace, agentic workflows writing into 3GTMS for load tracking, email triage, and freight-bill exception matching — together returning 160+ operational and 100+ back-office hours every week. I also lead an agent-authored development program producing 80% of the team's pull requests at 90%+ adoption, backed by hundreds of golden datasets and regression tests as the safety net.
Outside of NFI I build my own AI products end to end. CorkAI is an AI-native wine discovery platform I founded — the place where my engineering work and my Level Four (Advanced) sommelier training meet. Prodigy is a personal engineering project: a 95-function tool-using AI orchestrator built around a custom Model Context Protocol server. I'm currently completing an M.S. in Artificial Intelligence at the University of Colorado Boulder.
Experience
Where the work happened.
FEB 2016 — PRESENT
NFI Industries
Philadelphia, PA
Senior Director of Engineering · previously Director of Supply Chain Engineering, Director of Professional Services, Director of Product Management · Product Owner, Transportation Digital Twin (since 2018)
- Built NFI's consulting and digital twin practice from $0 to seven-figure revenue, leading 50+ network-design engagements at 10–15% average savings and supporting ~$700M in customer sales; together this work delivered $200M+ in total supply chain savings at NFI. Cut RFP and continuous-improvement turnaround from 6 weeks to 1.
- Own the Transportation Digital Twin as Product Owner since 2018, scaling it from a $500K pilot into a production B2B SaaS decision platform now modeling roughly $10M in annual savings across 10 enterprise accounts.
- Own an eight-figure vendor and partner portfolio across AI, optimization, GitHub, and staffing partners, leading evaluation and negotiation alongside NFI's President of Transportation Management, VP of Innovation, and VP of Procurement.
- Led a six-month phased production migration of NFI's optimization platform from Supply Chain Guru Cloud to Nextmv against a hard license-expiration deadline — including a full Databricks deployment — running live production and validation workloads in parallel on the same infrastructure, then cutting over account by account across 10 enterprise accounts with no service interruption.
- Ship AI/ML into live transactional systems — an XGBoost tender-acceptance model writing to Navitrace with tendering executing in 3GTMS, and n8n-orchestrated agentic workflows writing into 3GTMS for load tracking, email triage, and freight-bill exception matching — saving 160+ operational and 100+ back-office hours per week.
- Built an internal platform and infrastructure organization from 0 to 20+ across the US, Colombia, Brazil, Panama, India, and Europe, and was part of the leadership team for a 50+ engineer offshore staff-augmentation engagement with Zeb Group (about $1M a year, focused on AI/ML), where many of the engineers reported to me directly. Hired 30+ engineers directly and promoted several into management — career frameworks later formalized into a cross-divisional rotation program approved by the CEO/CFO.
- Lead an agent-authored development program across the engineering team — 90%+ team-wide AI adoption and 80% of PRs agent-authored, human-reviewed — backed by hundreds of golden datasets and regression tests as the quality safety net.
- Cut new-account onboarding from 3 months to 2 weeks and raised deployment cadence from monthly to several times per week via DORA metrics, Architecture Decision Records, and disciplined Agile/Scrum practice; direct team OKRs spanning feature delivery, roadmap prioritization, and technical-debt reduction.
- Launched Mira, an internal AI assistant, and more than doubled monthly usage within 90 days — owning the knowledge base, RAG chunking/retrieval, and RLHF, then extending it into a self-service portal with natural-language-to-SQL, automated reporting, and SOP knowledge automation.
- Lead a machine-learning team building production models on Databricks — XGBoost tender-acceptance prediction and clustering-based partition selection — including an eight-year carrier-tender study training a three-class classifier at ~90% accuracy.
- Serve as NFI's governance partner for AI and optimization tooling with IT, DevOps, and InfoSec alongside the CTO — consolidating a fragmented vendor landscape, running gap analyses, building standardized evaluation rubrics, and designing tool approval workflows.
CorkAI — AI-Native Wine Discovery & Recommendation Platform
Founder & CEO- Built a hybrid recommendation engine blending a 495-varietal knowledge base with LLM-based bottle-specific analysis, versioned and cache-invalidated for consistency across app, chatbot, and marketplace surfaces.
- Directed a multi-provider AI strategy (OpenAI GPT-4o family, Google Gemini 2.5 Flash) with cost-driven model routing and escalation, plus a grounded computer-vision bottle scanner with anti-hallucination rules and human confirmation gating.
- Directed a canonical wine-data platform resolving every user action into a confidence-weighted master record — source-weighted merge logic, append-only provenance, and an automated similarity-based dedup sweeper — shipped behind feature flags via phased, risk-managed rollout.
Prodigy — Personal AI Engineering Project
Creator & Product Lead- Independently engineered a ~95-function tool-using AI orchestrator — bucket-based context routing across 13 domains, enforced tool budgets, multi-specialist fan-out — around a custom bidirectional Model Context Protocol (MCP) server/client.
- Applied rigorous LLMOps evaluation (golden datasets, holdout sets, cross-model consensus) and a human-approval-gated governance pattern for every AI-generated artifact.
JUL 2014 — FEB 2016
LLamasoft, Inc.
Ann Arbor, MI · later acquired by Coupa
Product Engineer · Supply Chain Design Consultant
- Delivered network design, transportation optimization, and predictive-analytics engagements for enterprise customers, building Tableau dashboards that translated model outputs into executive-ready recommendations.
- Coordinated Agile sprint planning and backlog management in Jira and Confluence, designing rapid model-building capabilities later commercialized as a professional-services offering.
- Prototyped AWS landing-zone templates and end-to-end model workflows; specified a predictive-analytics workflow and helped build a proof of concept.
Featured Work
Two products and a method, one habit: ship it, then prove it works.
How I actually get an AI system from a working demo to something safe to put in front of customers — the operating model, the evaluation discipline, the decision framing, and the product judgment behind it.
White paper · Product operating model
The Control Tower Method — roadmaps, gates and maturity for an AI-built product
How I move a product's controls from manual to autonomous: a roadmap control tower fed by the system of record, review gates that CI enforces, and an evidence-based L0–L5 maturity matrix that decides how much each agent is trusted to do on its own. I first set it up at NFI for the transportation digital twin, across 27 control areas. CorkAI runs on it today, built by multiple Claude Code agents (Opus 5.5, medium effort) under one product owner.
27
control areas scored on the maturity ladder for NFI's digital twin
L0 → L5
from manual controls to controlled, autonomous agents
~1,150
automated tests gating every change
Prodigy · Creator & Product Lead
Prodigy & Maia — teaching an assistant to understand a whole sentence
Prodigy is an AI platform I designed and built myself — partly to organize my own work, mostly to take an AI capability all the way from idea to something dependable enough to trust daily, and to learn MCP, LLMOps, and honest evaluation by shipping something that breaks. Maia is its assistant, and Maia kept dropping requests, inventing details, and confirming things it hadn't done. So I built the planner underneath it, then the evaluation apparatus to prove whether it actually worked — which turned out to be the harder and more valuable half. The linked brief is the leadership case for switching it on, written to read for technical and non-technical audiences alike.
68% → 99%
task accuracy, honest baseline to shipping build
56%
of failures fixed by product rules, not a bigger model
86%
below the cost ceiling set before testing began
Founder & CEO · consumer AI product
CorkAI — AI-Native Wine Discovery & Recommendation Platform
An independent product I built from the ground up, and the place my engineering work meets my Level Four (Advanced) sommelier training. A hybrid recommendation engine blends a 495-varietal knowledge base with LLM-based, bottle-specific analysis — versioned and cache-invalidated for consistency across the app, chatbot, and marketplace. A grounded computer-vision scanner reads a label in seconds, with anti-hallucination rules and human confirmation before anything is trusted. Underneath, a canonical wine-data platform resolves every user action into a confidence-weighted master record with append-only provenance and an automated dedup sweeper, shipped behind feature flags through a phased rollout.
495
varietals in the core knowledge base
Level 4
Advanced Sommelier, plus 7 more wine credentials behind the domain model
0 → CEO
idea to founder-led product, built solo
Skills
What I actually work in.
AI / ML in Production
Agentic AIAutonomous Process Execution Management (APEM)Agent-authored developmentLLMOpsRAGRLHF
Model Context Protocol (MCP)Multi-model evaluation
Applied ML — regression, classification, decision trees, clustering
Deep learning & neural networks — CNNs, RNNs, LSTMsComputer vision
Human-in-the-loop governance
Cloud, Infrastructure & Migration
Live production replatformingPhased cutover & parallel-run validation
DatabricksKubernetesDocker
CI/CD — GitHub Actions, Jenkins, OpenShiftGCPAWS
Engineering Leadership
Org buildingNearshore & offshore deliveryManaging managers
DORA metricsOKRsArchitecture Decision RecordsAgile / Scrum
Multi-million-dollar vendor managementP&L ownership
AI Governance & Vendor Strategy
Vendor consolidation & gap analysisEvaluation rubricsTool approval workflows
Cross-functional governance (CTO / IT / DevOps / InfoSec)
Programming & Tooling
Python — TensorFlow, Keras, PyTorch, XGBoost, Scikit-learn, SciPySQLGo
JiraConfluenceTableauPower BIn8n
Logistics Technology
TMS — 3GTMS, Aljex, Descartes, TMWWMS — Manhattan, Red Prairie
Optimization — Nextmv, Supply Chain GuruERP — SAP, OracleNavitraceTransporeonCoupa
Education
Formal training, kept current.
M.S., Artificial Intelligence
University of Colorado Boulder
In progress · expected 2028
Specialization pathways
Machine Learning
Probability & Statistics
Introduction to Artificial Intelligence
Reinforcement Learning
AI Ethics
Software Architecture for Big Data
Generative AI
Computer Vision
Deep Learning
Recommender Systems
B.S., Logistics, Materials & Supply Chain Management
University of Arkansas
Cum Laude · 2014
Spoken Languages
English
Spanish
Portuguese
French
Italian
Operational Leadership
- DORA Metrics & Agile Delivery Discipline — self-directed implementationNFI Industries
- Cross-Divisional Rotation Program Design — approved by CEO/CFONFI Industries
Certifications
39 certifications, each one verifiable.
Grouped by relevance, from machine learning and AI ethics through generative AI, product management, applied AI, and wine. Search by name, issuer, or topic, re-sort the list, or open any certificate at full size.
Showing all 39 certifications
Machine Learning & Data Science
5AI Ethics & Governance
5Generative & Agentic AI
12ChatGPT: Master Free AI Tools to Supercharge Productivity Specialization
Vanderbilt University
ES3R9T8311WUVerify
Product & Project Management
4ChatGPT for Project Management - Leveraging AI for Success Specialization
Vanderbilt University
FT3GP6I2FAZ3Verify
ChatGPT for Project Management: Insight, Planning, & Success
Vanderbilt University
379RFH7G4600Verify
ChatGPT for Project Management: Execution, Tracking, Success
Vanderbilt University
MGJFGLFT5Q0TVerify
Applied AI: Security & Marketing
5Wine & Sommelier
8The domain expertise behind CorkAI’s recommendation engine and tasting vocabulary.
No certifications match that search.
Writing
Published writing.
Industry pieces on supply chain modeling and digital twins, plus the wine education library I research and publish on CorkAI — where the sommelier training does the work.
Selected Publications
NFI Industries · 2020
Digital Twin — The Key for Proactive Supply Chain Planning
How a virtual replica of a supply chain lets a company test scenarios and decide proactively, instead of reacting to disruption after it lands. Co-authored with NFI's senior scientist.
Read the article↗
NFI Industries · 2022
Using the Digital Twin: 4 Supply Chain Scenarios That Generate Savings
Four concrete modeling scenarios — and the savings each one surfaces — drawn from live enterprise network-design engagements.
Read the article↗
Logistics Tech Outlook · CXO Insights
The Smarter Way to Model Supply Chains
Why most network models answer the wrong question, and what changes when you model decisions rather than snapshots.
Read the article↗
Logistics Tech Outlook · 2019
Four Must-Have Skills for Supply Chain Analytics
What separates analysts who produce dashboards from analysts who change decisions. Written as a CXO Insights column.
Read the article↗
Wine Education — CorkAI
Featured guide · Wine basics
How to Pick Wine: A CorkType Guide
Choosing wine by personality rather than by grape. Introduces the "Pick This, Not That" rule for finding bottles you'll actually love, every time.
6 min readRead ↗
Wine basics
Where to Buy Wine: The Ultimate Guide
How smarter wine shopping changes the experience — from curated clubs to personalized picks matched to your taste.
12 min readRead ↗
Wine basics
What Are Appellations
Why place matters more than grape — how to read a label and which specific regions are worth knowing.
8 min readRead ↗
Wine basics
Is Riesling Sweet or Dry? A Buyer's Guide to Every Style
Riesling runs bone-dry to dessert-rich, and one label trick tells you which. Includes an ABV cheat sheet.
8 min readRead ↗
Wine basics
Vitis Vinifera
From ancient Transcaucasus vineyards to modern regions — how one grape species produces infinite expressions.
5 min readRead ↗
Wine basics
Your Wine Personality: Why CorkTypes Make Wine Simple
A two-question framework for understanding which wines naturally appeal to you. No prior wine knowledge required.
5 min readRead ↗
Tutorial
The CorkAI Wine Scanner: How It Works
Point a phone at a label — or a whole row of them — and get producer, region, vintage, taste profile, and pairings back.
7 min readRead ↗
Tutorial
CorkChat: Your AI Wine Conversation Partner
Having a natural conversation about wine with an AI sommelier that remembers your preferences and sharpens over time.
10 min readRead ↗
Tutorial
Master the Sommelier's Pick: Your Personal Wine Pairing Assistant
Using CorkAI's pairing engine to find matches for a specific dish, occasion, or mood.
8 min readRead ↗
Tutorial
Mastering Your Wine Cellar: A Complete Guide
Building and managing a digital wine collection — from scanning labels to tracking what you've opened.
10 min readRead ↗