Pre-Conference Workshop · FOPAM 2026

Foundations to Frontier: Machine Learning & AI for the Process Industries

From Classical Methods to LLMs and Agents, to What Works on the Plant Floor

Dates
Sunday–Monday, July 26–27, 2026
Time
1:00 PM Sunday – 4:30 PM Monday
Location
Decatur, GA
  Register via FOPAM 2026   See full schedule
Design of ExperimentsMachine LearningDeep LearningLarge Language ModelsProcess IndustriesData AnalyticsHands-On
01 · Overview

Why this workshop, why now

This workshop offers process and product engineers, data scientists, and graduate students a single, coherent arc through machine learning and AI for the process industries: from the classical statistical foundations that have driven process and product development for decades, to today’s deep learning, large language models (LLMs), and agentic systems, and finally to what actually works on the plant floor.

Rather than a survey of disconnected methods, the workshop is built as three linked modules taught by leaders who span academia, pharmaceutical R&D, and chemicals manufacturing. Module 1 grounds participants in the experimental design, data, and statistical modeling that remain the bedrock of sound analytics. Module 2 builds from there to the modern methods reshaping the field. Module 3 closes the loop with hard-won industrial reality and hands-on datasets.

Participants leave understanding not just how these methods work, but when to reach for each and what it takes to deploy them responsibly.

  Why now?

The process industries (pharmaceuticals, chemicals, materials, energy) are in the middle of a genuine shift in how data and models are used. Classical design of experiments and multivariate statistics remain indispensable, yet deep learning, foundation models, and LLM-based agents are now entering process design, monitoring, and operations.

Practitioners face a widening gap between what is demonstrated in papers and what is dependable in a regulated, safety-critical plant. This workshop is timed to meet that gap head-on at FOPAM, the premier venue bridging process systems engineering and machine learning. It pairs the rigor of classical methods with an honest assessment of modern AI and the practitioner’s view of what survives contact with real manufacturing data.

02 · Learning outcomes

What you'll take away

After attending, participants will be able to:

1
Design sound experiments and apply classical statistical, regression, and dimensionality-reduction methods to process/product data.
2
Explain how deep learning, LLMs, and agents work, and critically evaluate claims about their capabilities and limits.
3
Recognize where modern AI adds real value versus where classical methods remain the right tool.
4
Anticipate the data-quality, deployment, and organizational challenges of putting ML/AI into industrial practice, informed by hands-on work on real datasets.
03 · Schedule

Three sessions over a day and a half

The workshop runs in the optional pre-conference slot, beginning Sunday afternoon and concluding Monday afternoon, ahead of the FOPAM Welcome Reception.

Module 1
Sunday, July 26
1:00 PM – 4:00 PM
Design of Experiments, Data Exploration Analysis, Classical Regression, and Active Learning
Jose E. Tabora · Bristol Myers Squibb
A hands-on lecture on applications of DOE to optimize unit operations or processes.
Module 2
Monday, July 27
9:00 AM – 12:00 PM
Modern ML/AI: Deep Learning, LLMs, and Agents
R. Bhushan Gopaluni · University of British Columbia
A grounded path from classical models to the methods defining the current frontier, framed for process applications.
Module 3
Monday, July 27
1:30 PM – 4:30 PM
AI/ML in Industry: A Journey Through Industrial Examples
Bea Braun · Dow
A journey through industrial examples and an opportunity to get your hands dirty.

Each 3-hour session includes a break. Schedule is provisional and subject to confirmation with the FOPAM organizers.

04 · Modules in detail

What you'll cover

Three linked modules that build on each other — from classical foundations to modern AI to industrial reality.

1
Module 1 Sunday, July 26 · 1:00 PM – 4:00 PM

Design of Experiments, Data Exploration Analysis, Classical Regression, and Active Learning

Jose E. Tabora, Bristol Myers Squibb

DOE Workflow 60 min
  • Hypothesis — Taylor Series Expansion
  • DOE Types — Fisher Information Matrix
  • Data Collection
  • Randomization
Data Exploration Analysis 30 min
  • Color maps, scatter plots, beeswarm plots
  • Dimensionality reduction: PCA, t-SNE
Regression Approaches 60 min
  • Generalized Linear Models (GLM)
  • Bayesian Optimization (BO)
Model Visualization and Decision Making 30 min
  • Interpreting model outputs
  • Translating models into actionable decisions
2
Module 2 Monday, July 27 · 9:00 AM – 12:00 PM

Modern ML/AI: Deep Learning, LLMs, and Agents

R. Bhushan Gopaluni, University of British Columbia

From Regression to Representation 50 min
  • Why representation learning matters for process data
  • Neural networks: architecture and training
  • Deep learning for time-series and process signals
Sequence Models, Transformers, and LLMs 50 min
  • From RNNs to attention mechanisms
  • How transformers and LLMs work
  • What LLMs can and cannot do for engineering tasks
LLMs as Agents 50 min
  • Reasoning, planning, and tool use
  • Retrieval-augmented generation (RAG)
  • Building agents for process engineering tasks
Limitations and Responsible Deployment 20 min
  • Hallucination and numerical reasoning failures
  • Safety considerations for industrial settings
  • When to use (and not use) LLM-based solutions
3
Module 3 Monday, July 27 · 1:30 PM – 4:30 PM

AI/ML in Industry: A Journey Through Industrial Examples

Bea Braun, Dow

What Makes 'Math to Money' Hard? 20 min
  • The gap between models and deployment
  • Common failure modes in industrial ML
Classical ML Still Does the Heavy Lifting 80 min
  • Hands-on: Formulations dataset
  • Hands-on: Industrial dataset with PLS modeling
  • Why classical methods remain essential
Modern Machine Learning in Action 40 min
  • Seeing opportunities for machine vision
  • Generative AI and agentic frameworks in industrial settings
Ingredients for Success 30 min
  • How to really make it work
  • Organizational and cultural factors
  • Discussion and Q&A

  What to bring

  •   A laptop (required for hands-on sessions in Modules 1 and 3).
  •   Ability to connect to the workshop Wi-Fi.

  What we provide

  •   Hands-on datasets: formulations and industrial process data.
  •   Jupyter notebooks and starter code.

  Prerequisites

Familiarity with basic engineering data analysis is helpful. No prior deep-learning or programming experience is required. Module 1 establishes the foundations and each module is self-contained while building on the last. Modules 1 and 3 include hands-on components, so participants should bring a laptop.

05 · Speakers & Organizers

Across academia, pharma, and chemicals

Leaders who span pharmaceutical R&D, chemicals manufacturing, and academic research.

Jose E. Tabora
Speaker

Jose E. Tabora

Bristol Myers Squibb
Senior Scientific Director, Chemical Process Development
  New Brunswick, NJ
  Sessions: Module 1

AIChE Fellow and recipient of the 2022 AIChE Industrial Research & Development Award, with 25 years in drug-substance development across Merck, Eli Lilly, and BMS. Expertise in design of experiments, statistical and mechanistic modeling, data analysis, and visualization. Has taught graduate courses in reactor design, reaction kinetics, and data science at Manhattan College and Stevens Institute of Technology. PhD, Chemical Engineering, University of Virginia.

R. Bhushan Gopaluni
Speaker

R. Bhushan Gopaluni

University of British Columbia
Professor, Chemical & Biological Engineering; Vice-Provost & AVP, Faculty Planning
  Vancouver, BC
  Sessions: Module 2

Directs the Data Analytics and Intelligent Systems (DAIS) Lab, focused on machine learning, reinforcement learning, and data analytics for the process industries. Author of 150+ refereed publications; recipient of the Killam Teaching Prize and the D.G. Fisher Award (Canadian Society for Chemical Engineering). Extensive experience organizing workshops at process-systems venues, including FOPAM 2023, ADCHEM, and DYCOPS.

Bea (Birgit) Braun
Speaker

Bea (Birgit) Braun

Dow
R&D Fellow, Digital Innovation; Principal Research Scientist, Data Science & Hybrid Modeling
  Freeport, TX
  Sessions: Module 3

Nearly 15 years at Dow leading the development and deployment of data-science solutions across R&D and manufacturing, with more than a decade applying AI/ML to chemical manufacturing and materials research. Six Sigma Black Belt with 15 journal papers and 70+ internal publications. Serves on the FOPAM 2026 organizing committee. PhD, Chemical Engineering, Colorado School of Mines.

06 · Target audience

Who this is for

  •   Process and product engineers and R&D scientists in pharma, chemicals, materials, and energy.
  •   Data scientists and analytics practitioners working with process/product data.
  •   Academic researchers and graduate students in process systems engineering and applied ML.

Expected attendance: 25–50 participants spanning academia and industry.

07 · About FOPAM

Foundations of Process/Product Analytics

FOPAM 2026 is the premier conference bridging process systems engineering and machine learning, organized by CACHE. The conference runs July 26–30, 2026 at the Courtyard by Marriott Atlanta Decatur Downtown/Emory in Decatur, GA.

Learn more about FOPAM 2026 →

08 · Registration

Join us in Atlanta

Workshop registration is handled as an optional add-on through the FOPAM 2026 conference site.

  Sunday–Monday, July 26–27, 2026
  1:00 PM Sunday – 4:30 PM Monday
  Courtyard by Marriott Atlanta Decatur Downtown/Emory
  FOPAM 2026 Registration   View schedule

Questions? Contact bhushan.gopaluni@ubc.ca

This workshop is part of FOPAM 2026, organized by CACHE. Schedule and details are provisional, pending final confirmation with the FOPAM organizing committee.