#40 Soren Little: Optimization Inside the Database

What if your database could make decisions? In this episode, Vijay Mehrotra and Michael Watson talk with Soren Little, co-founder of OSM, about making optimization a built-in part of the data stack. Databases already have SELECT for descriptive analytics and PREDICT for forecasting. OSM Data argues the missing keyword is DECIDE.The conversation covers why predict-then-optimize falls short, how scale makes stochastic optimization practical, and the challenge of trading battery storage in volatile energy markets. It ends with what Soren calls one of the defining questions of the next five years: scheduling AI compute without destabilizing the power grid.What you’ll learn- Why optimization should be a first-class citizen in the database- How separating what you want to solve from how to solve it mirrors what SQL did for data retrieval- How a system can spot a cheaper problem type, such as bipartite matching, before reaching for a general MIP solver- Why data scientists can move from dashboards and forecasts to recommended actions- Where predict-then-optimize breaks down, and how large-scale methods could make stochastic and robust optimization practical- The SELECT → PREDICT → DECIDE progression in database analytics- Why data centers, batteries and the grid can be treated as one supply chainTimestamps0:00 - Preview & Introduction1:17 - Meet Soren Little2:45 - What OSM does4:10 - The founding team: Soren, Jakob, and Matteo7:25 - The Decision House and its Lakehouse inspiration11:44 - Separating the business problem from the problem-solution method15:40 - Picking the best formulation automatically: the GPU scheduling example20:45 - Bringing optimization to the VLDB database community22:35 Beyond the dashboard: data scientists as decision makers24:45 - Handling uncertainty and the limits of predict-then-optimize28:40 - SELECT, PREDICT, DECIDE31:00 - Real-world use cases: airline scheduling and battery storage36:12 - OSM's vision: AI compute, energy, and the gridFollow the showApple: ⁠https://podcasts.apple.com/in/podcast/the-decision-intelligence-lab/id1811085064⁠Spotify: ⁠https://open.spotify.com/show/0lFoAVKqJHTYSZNpeN61ou?si=0ae973aab0174b3b⁠Connect with guestSoren Little: ⁠https://www.linkedin.com/in/sorenlittle/OSM: https://www.osm-data.com/Connect with hostsProf. Vijay Mehrotra (University of San Francisco): ⁠https://www.linkedin.com/in/vijay-mehrotra-ba9498/⁠Prof. Michael Watson (Northwestern University): ⁠https://www.linkedin.com/in/michael-watson-07600a1⁠About the podcastThe Decision Intelligence Lab podcast delivers real-world insights for data professionals, business leaders, and anyone seeking to leverage data & AI for smarter decision-making & successful business outcomes.For business inquiries, email at ⁠[email protected]⁠

Make better business decisions with data and AI—subscribe to The Decision Intelligence Lab Newsletter at ⁠⁠https://decisionintelligencelab.substack.com/⁠⁠.What happens after your ML model makes a prediction? When Jerry Yurchisin asks that at data science conferences, the answer is usually "I don't know." In this episode, Gurobi's senior data science strategist joins Vijay and Mike to talk about the tool that turns predictions into decisions: optimization.They cover why you don't need a PhD to deliver real value, why solvers never forget a constraint the way an LLM can, how GenAI helps experts model faster (while confidently getting things wrong) and why finance may be leaving optimization value on the table.If you're wondering what comes after prediction, this one's worth listening to.Timestamps: 0:00 - Preview 0:42 - Meet Jerry Yurchisin4:25 - What a Senior Data Science Strategist does at Gurobi5:45 - Spreading the word beyond OR conferences (Gartner, Databricks, ODSC)7:00 - Statistics + OR at UNC9:30 - The pitch to ML practitioners: "What happens next?"11:11 - Optimization as a career differentiator13:40 - Is optimization too math-heavy?17:42 - "I tried MIP ten years ago and it was too slow" - try again21:02 - When "optimization" means different things to ML people22:35 - Building Optimization 101 for data scientists25:10 - Do executives confuse data science and optimization?26:30 - Teaching with realistic (but solvable) examples29:10 - Applying GenAI and optimization together32:10 - Advice for the OR community: join the AI conversation35:40 - Hard constraints vs. context windows37:15 - Underutilized frontier: optimization in finance39:20 - Learn more at gurobi.com/learnFollow the showApple: ⁠https://podcasts.apple.com/in/podcast/the-decision-intelligence-lab/id1811085064⁠Spotify: ⁠https://open.spotify.com/show/0lFoAVKqJHTYSZNpeN61ou?si=0ae973aab0174b3b⁠Connect with guest- Jerry's LinkedIn: https://www.linkedin.com/in/mathwithjerome/- Gurobi: https://www.gurobi.com- Gurobi learning resources: https://www.gurobi.com/learnConnect with hosts- Prof. Vijay Mehrotra (University of San Francisco): ⁠https://www.linkedin.com/in/vijay-mehrotra-ba9498/⁠- Prof. Michael Watson (Northwestern University): ⁠https://www.linkedin.com/in/michael-watson-07600a1⁠About the podcastThe Decision Intelligence Lab podcast delivers real-world insights for data professionals, business leaders, and anyone seeking to leverage data & AI for smarter decision-making & successful business outcomes.For business inquiries, email at ⁠[email protected]⁠

Make better business decisions with data and AI—subscribe to The Decision Intelligence Lab Newsletter at ⁠⁠https://decisionintelligencelab.substack.com/⁠⁠.What does it take to lead when the stakes are literally life and death? In this episode, Vijay and Mike sit down with Dr. Rich Staats — retired U.S. Army Major General, MIT Operations Research PhD, Hertz Fellow, and former Commanding General of the U.S. Army Reserve's 75th Innovation Command. Rich shares the three leadership principles he published in every command he held, the story of dining across the table from a man who threatened to kill him — and the one sentence that defused it — and why operations research sat at the center of every decision in his 41-year career. The conversation covers how much data a leader really needs, his "law of risk", lessons from running MITRE's OR department on briefing senior decision makers, and a candid take on generative AI: where it genuinely helps, and why he worries young analysts are losing their analytic muscle memory. Timestamps0:00 - Preview0:47 - Meet Dr. Rich Staats2:00 - How Rich and Mike met at an INFORMS conference3:05 - What ties a career across combat, think tanks, and startups together8:25 - "If it hadn't been for the war, we'd have retired several ranks earlier"11:03 - The Iraq dinner story: "I could just kill you now"14:14 - Double down on strengths or fix your weaknesses?15:50 - Innovation in the military & the 75th Innovation Command19:43 - OR at the center of 41 years of decisions22:30 - How often did he actually build a model? 24:06 - The Red Pen Award: rewarding the best mistakes25:23 - How much data do you need? The 80/20 lesson28:10 - Advising senior executives: decision space, ranges, and the ORmadillos32:40 - Generative AI: biggest hopes and red flags37:30 - IP ownership and liability: the next 15–20 year challenge38:58 - Where to find Rich What You'll Learn - The three leadership principles (Perrin, Stafford, Peterson) that work in combat units, think tanks, and startups alike- Why respect and relationships (not command-and-control) kept Rich alive in Iraq- How an OR mindset can shrink 2-hour meetings to 15 minutes - How much data you actually need before deciding and why waiting for certainty is its own risk- How to brief analysis to senior senior decision makers: decision space, ranges, and the value of information- Where generative AI helps and how it can quietly erode analytic "muscle memory" and creativityFollow the showApple: ⁠https://podcasts.apple.com/in/podcast/the-decision-intelligence-lab/id1811085064⁠Spotify: ⁠https://open.spotify.com/show/0lFoAVKqJHTYSZNpeN61ou?si=0ae973aab0174b3b⁠Connect with guest- Rich's Website: https://zhalindor.com/- Rich's Music: https://zhalindor.com/dreams- Rich's Books: https://zhalindor.com/booksConnect with hosts- Prof. Vijay Mehrotra (University of San Francisco): ⁠https://www.linkedin.com/in/vijay-mehrotra-ba9498/⁠- Prof. Michael Watson (Northwestern University): ⁠https://www.linkedin.com/in/michael-watson-07600a1⁠About the podcastThe Decision Intelligence Lab podcast delivers real-world insights for data professionals, business leaders, and anyone seeking to leverage data & AI for smarter decision-making & successful business outcomes.For business inquiries, email at ⁠[email protected]⁠

Make better business decisions with data and AI—subscribe to The Decision Intelligence Lab Newsletter at ⁠⁠https://decisionintelligencelab.substack.com/⁠⁠.Jeff Camm, Professor and Inmar Presidential Chair in Analytics at Wake Forest University School of Business joins Vijay and Mike to explain why so many data science projects fail to deliver ROI, and what his analysis of over 1.4 million job postings reveals about the skills that actually drive decisions.He shares how he built Wake Forest's MSBA around "the bookends" (problem framing and influence), his blueprint for a new undergrad core(Artificial Intelligence, Business Intelligence, Decision Intelligence) and why managers want choices, not answers. check against transcript for factual accuracyIf your models aren't changing decisions, this one's for you.What You'll Learn- The measurable skill-set differences between data science and decision science roles, based on a decade of job posting data- Why data science projects fail and why the reasons echo OR project failures from decades past- How to design an analytics program around "the bookends": problem framing and influence- How to actually teach problem framing (experiential learning, real client messes, no data up front)- A three-course blueprint for the undergrad business core: Artificial Intelligence, Business Intelligence, Decision Intelligence- Why prescriptive analytics should deliver families of solutions, not single answers- Why you should ask an LLM for choices, not answers and what that means for teaching analytics in the AI eraTimestamps0:00 - Preview1:01 - Meet Jeff Camm1:40 - Why "decision scientist," not "data scientist"4:20 - The research outcome: academia versus 1.4M job ads 9:10 - Why data projects fail12:50 - Building Wake Forest's Masters in Business Analytics from scratch14:42 - The bookends: problem framing and influencing18:22 - Teaching problem framing: the practicum and "Mess to Model"20:35 - Team charters and learning to work in teams22:50 - A new undergrad core: AI → BI → Decision Intelligence29:10 - Problem-centric teaching and just-in-time technique33:30 - "Nobody likes to be told what to do": choices, not answers34:50 - LLMs, families of solutions, and the risk/return trade-off38:40 - Jeff's papers and where to learn moreResources: - "Data Science and Decision Science Skills: Are They Different and Does It Matter?" (Camm, Fry & Shafer, Harvard Data Science Review, Summer 2025): ⁠https://hdsr.mitpress.mit.edu/pub/9ir6e1j6/release/1⁠- The INFORMS Journal on Applied Analytics: ⁠https://pubsonline.informs.org/authored-by/Camm/Jeffrey+D⁠- "Stop Modeling Just the Data, Start Modeling Decisions" (forthcoming in MIT Sloan Management Review digital at time of publishing)Follow the showApple: ⁠https://podcasts.apple.com/in/podcast/the-decision-intelligence-lab/id1811085064⁠Spotify: ⁠https://open.spotify.com/show/0lFoAVKqJHTYSZNpeN61ou?si=0ae973aab0174b3b⁠Connect with guest- Jeff Camm: ⁠https://www.linkedin.com/in/jeff-camm-395b366/Connect with hosts- Prof. Vijay Mehrotra (University of San Francisco): ⁠https://www.linkedin.com/in/vijay-mehrotra-ba9498/⁠- Prof. Michael Watson (Northwestern University): ⁠https://www.linkedin.com/in/michael-watson-07600a1⁠About the podcastThe Decision Intelligence Lab podcast delivers real-world insights for data professionals, business leaders, and anyone seeking to leverage data & AI for smarter decision-making & successful business outcomes.For business inquiries, email at ⁠[email protected]⁠

Make better business decisions with data and AI—subscribe to The Decision Intelligence Lab Newsletter at ⁠⁠https://decisionintelligencelab.substack.com/⁠⁠.What does it actually take to run a marketplace across thousands of micro-markets in real time? Cyrus Safaie, Former Director of Engineering and AI/ML at DoorDash joins Mike and Vijay to pull back the curtain on how DoorDash balances supply and demand, pays dashers, and predicts delivery times at massive scale. Cyrus shares why the company started with spreadsheets instead of algorithms, how a surprise Netflix boxing match drove a Super Bowl-sized demand spike with zero warning, and why the best automated systems treat human judgment as an input rather than an override. He also previews his new venture: using AI to build operationally heavy companies run by a single domain expert or tiny team.Timestamps0:00 - Preview1:01 - Meet Cyrus Safaie2:00 - What people underestimate about DoorDash: hyper-local markets3:28 - Scaling by doing things that don't scale6:48 - Optimizing a three-sided marketplace & the trade-offs11:20 - Who resolves conflicts between dasher, merchant, and consumer teams13:00 - Balancing supply/demand and dasher incentives in real time16:48 - Reactive Real-time decisions and the Netflix Tyson fight21:30 - Human input vs. human override: how to automate the right way27:27 - Fixing ETAs by predicting your own model's errors29:35 - Cyrus's new venture: AI-native operationally heavy companies33:38 - Where the human stays in the loop: one-expert companies36:36 - Advice for students: curiosity and "torturing your brain"What You'll Learn- Why DoorDash is really thousands of tiny, semi-isolated markets and why it optimizes locally before globally- How DoorDash scaled by starting with spreadsheets and manual judgment before building algorithms- How a three-sided marketplace balances dashers, merchants, and consumers- The difference between proactive and reactive dasher incentives, and how real-time mobilization works- Why human input into automated systems beats human overrides of them- How DoorDash improved ETAs by building models that predict their own errors- Cyrus's vision for AI-native, operationally heavy companies run by a single domain expert (or tiny team) plus an AI operating systemFollow the showApple: ⁠https://podcasts.apple.com/in/podcast/the-decision-intelligence-lab/id1811085064⁠Spotify: ⁠https://open.spotify.com/show/0lFoAVKqJHTYSZNpeN61ou?si=0ae973aab0174b3b⁠Connect with guestCyrus Safaie: ⁠https://www.linkedin.com/in/hsafaie/Connect with hostsProf. Vijay Mehrotra (University of San Francisco): ⁠https://www.linkedin.com/in/vijay-mehrotra-ba9498/⁠Prof. Michael Watson (Northwestern University): ⁠https://www.linkedin.com/in/michael-watson-07600a1⁠About the podcastThe Decision Intelligence Lab podcast delivers real-world insights for data professionals, business leaders, and anyone seeking to leverage data & AI for smarter decision-making & successful business outcomes.For business inquiries, email at ⁠[email protected]⁠

Make better business decisions with data and AI—subscribe to The Decision Intelligence Lab Newsletter at ⁠⁠https://decisionintelligencelab.substack.com/⁠⁠.Airline veteran William Swelbar joins Vijay Mehrotra and Mike Watson for a tour through nearly 50 years of airline economics(from 1978 deregulation to today's premiumization era).William Swelbar's path is one of a kind: flight attendant at North Central in 1979, local union president in Detroit within a few years, and part of a 1982 employee coalition that had all but raised the $400 million it needed in a bid to buy Republic Airlines.The conversation digs into the analytics behind the industry: why hub-and-spoke networks turn 10 routes into 230 sellable city pairs, how Pan Am's $99 cabin became an early pricing wake-up call, why "capacity is a bad drug," and how Delta and United decommoditized flying through cabin segmentation, basic economy, and credit card revenue - while the end of labor arbitrage killed the low-cost carrier era.Timestamps0:00 - Preview0:44 - Vijay's Caddy Master Turned Airline Veteran5:10 - What deregulation actually changed8:00 - The overnight flood of new entrants9:30 - Pan Am's $99 cabin and ~120 bankruptcies11:43 - The fall of TWA and Pan Am; American's innovations 14:23 - Network design: Hub-and-spoke vs. point-to-point16:32 - Southwest and the Southwest Effect19:40 - High-speed rail: too late for the US?22:15 - Capacity is a "bad drug"25:15 - How Delta and United segmented the cabins & won during COVID29:05 - Airfares never covered the cost of flying30:45 - Basic economy as a weapon against Spirit and Frontier33:50 - The end of the value airline sector36:55 - Wrap-upFollow the showApple: ⁠https://podcasts.apple.com/in/podcast/the-decision-intelligence-lab/id1811085064⁠Spotify: ⁠https://open.spotify.com/show/0lFoAVKqJHTYSZNpeN61ou?si=0ae973aab0174b3b⁠Connect with guestWilliam Swelbar: ⁠https://swelbar.substack.com/Connect with hosts- Prof. Vijay Mehrotra (University of San Francisco): ⁠https://www.linkedin.com/in/vijay-mehrotra-ba9498/⁠- Prof. Michael Watson (Northwestern University): ⁠https://www.linkedin.com/in/michael-watson-07600a1⁠About the podcastThe Decision Intelligence Lab podcast delivers real-world insights for data professionals, business leaders, and anyone seeking to leverage data & AI for smarter decision-making & successful business outcomes.For business inquiries, email at ⁠[email protected]⁠

Make better business decisions with data and AI—subscribe to The Decision Intelligence Lab Newsletter at ⁠⁠https://decisionintelligencelab.substack.com/⁠⁠.Geoffrey De Smet built a scheduling solver as a side project and ended up with a NASA supplier, big telcos, and pest control companies running on it.The Timefold co-founder and OptaPlanner creator joins Vijay and Mike to talk field service routing: hundreds of technicians, tens of thousands of jobs, and the chaos of a real day in the field. They also discuss why plans are useless but planning is essential, why a 90% feasible schedule is 100% useless, why operators reject schedules they can't interrogate and what it took to leave a steady job when his life's work became roadkill after the IBM–Red Hat acquisition.If you've ever wondered why the world still runs on scheduling spreadsheets, this one's for you.Chapters0:00 - Preview & Introduction1:00 - Meet Geoffrey De Smet, Co-founder Timefold1:20 - What is field service routing?3:05 - Non-disruptive replanning and uncertainty10:25 - What-if simulations11:35 - Who uses Timefold Users14:45 - The telecom case: ROI and trade-offs of optimization18:05 - Spreadsheets and scheduling problems19:10 - From OptaPlanner to Timefold: The Origin story26:23 - Customising the model28:20 - Explainability deep dive30:28 - Where LLMs fit 31:50 - Trust, feasibility, constraints and the operator's never-ending world35:23 - The Leap: leaving a steady job, six months of burning savings38:07 - Go-to-market challenges40:40 - Closing thoughtsFollow the showApple: ⁠https://podcasts.apple.com/in/podcast/the-decision-intelligence-lab/id1811085064⁠Spotify: ⁠https://open.spotify.com/show/0lFoAVKqJHTYSZNpeN61ou?si=0ae973aab0174b3b⁠Connect with guestGeoffrey De Smet: ⁠https://www.linkedin.com/in/ge0ffrey/Timefold: https://timefold.aiTimefold Solver: https://timefold.ai/solverConnect with hostsProf. Vijay Mehrotra (University of San Francisco): ⁠https://www.linkedin.com/in/vijay-mehrotra-ba9498/⁠Prof. Michael Watson (Northwestern University): ⁠https://www.linkedin.com/in/michael-watson-07600a1⁠About the podcastThe Decision Intelligence Lab podcast delivers real-world insights for data professionals, business leaders, and anyone seeking to leverage data & AI for smarter decision-making & successful business outcomes.For business inquiries, email at ⁠[email protected]⁠

. Make better business decisions with data and AI—subscribe to The Decision Intelligence Lab Newsletter at ⁠⁠https://decisionintelligencelab.substack.com/⁠⁠.Optimization powers decisions in everything from healthcare to logistics, but for most practitioners it stays a "magical box": powerful, opaque, and locked behind PhD-level expertise. So what actually gets in the way of putting these models to work?In this episode, hosts Vijay Mehrotra and Michael Watson sit down with Stanford's Postdoctoral Researcher Connor Lawless and Madeleine Udell (Assistant Professor at Management Science and Engineering Department, Stanford University) to unpack their paper "It Was a Magical Box": Understanding Practitioner Workflows and Needs in Optimization. Based on interviews with 15 optimization model developers, the research uncovers a surprising truth: the hardest part of operations research usually isn't the math. It's the people, the data, and the endless back-and-forth.We dig into the six-stage workflow of building an optimization model, why nearly every project becomes an iterative "flywheel," why machine learning feels so much easier than OR, when "good enough" beats provably optimal solutions, and how LLMs might finally close the accessibility gap. Connor also shares what's next as he joins Percepta to work on the "last mile" of analytics.Whether you build models for a living or just wonder why so many great models never make it into the real world, this conversation will change how you think about optimization in practice.The Research Paper: "It Was a Magical Box": Understanding Practitioner Workflows and Needs in Optimization (2025) - https://arxiv.org/abs/2509.16402Timestamps0:00 - Preview & Introduction0:57 - Meet Connor Lawless and Madeleine Udell2:04 - OR's accessibility problem vs. ML3:07 - The six stages of building an optimization model5:38 - Iteration as a flywheel: the "99% of the time" finding7:45 - Why is ML so much easier than OR? 10:45 - The role of visualization and pattern recognition12:00 - Optimization as a distribution-shift problem13:24 - The big surprise: the human bottleneck, not the math15:32 - Implications for teaching in the age of AI18:00 - Handling uncertainty: data-scarce vs. data-rich problems20:00 - Solver friction: Gurobi, CPLEX, and parameter tuning22:20 - "Good enough" beats optimal24:16 - Practitioner innovations: data and constraint validators26:00 - Separating data from the model27:37 - Preventing silent wrong answers29:00 - Documentation, debugging, and LLMs as intermediaries30:30 - Connor's next chapter: Percepta and the "last mile" of analytics32:00 - If not us, then who?Follow the showApple: ⁠https://podcasts.apple.com/in/podcast/the-decision-intelligence-lab/id1811085064⁠Spotify: ⁠https://open.spotify.com/show/0lFoAVKqJHTYSZNpeN61ou?si=0ae973aab0174b3b⁠Connect with guest- Connor Lawless (Postdoctoral researcher, Stanford University): https://www.linkedin.com/in/connorlawless/- Madeleine Udell (Assistant Professor, Management Science & Engineering, Stanford University): https://www.linkedin.com/in/madeleine-udell/Connect with hosts- Prof. Vijay Mehrotra (University of San Francisco): ⁠https://www.linkedin.com/in/vijay-mehrotra-ba9498/⁠- Prof. Michael Watson (Northwestern University): ⁠https://www.linkedin.com/in/michael-watson-07600a1⁠About the podcastThe Decision Intelligence Lab podcast delivers real-world insights for data professionals, business leaders, and anyone seeking to leverage data & AI for smarter decision-making & successful business outcomes.For business inquiries, email at ⁠[email protected]⁠

Make better business decisions with data and AI—subscribe to The Decision Intelligence Lab Newsletter at ⁠⁠https://decisionintelligencelab.substack.com/⁠⁠.What if the biggest mistake organizations make isn't in their data or their AI, but in never actually defining the decision they're trying to make?In this episode of the Decision Intelligence Lab, hosts Vijay Mehrotra and Michael Watson sit down with James Taylor, the author of Digital Decisioning, and Executive Partner at Blue Polaris. James traces the idea back 25+ years to a simple observation in financial services: fraud, origination, and collections systems all ran on the same technology stack yet were treated as entirely separate things. Naming that pattern changed how a whole industry approaches automation.The conversation digs into why decisions should be isolated as their own unit of work, why you should aim to automate everything to discover the parts that truly need people, and how a disability-claims team once saved 15 months of work simply by understanding the decision before building the model. James also shares where decisioning is going next, why he partners so closely with IBM, and his three durable principles for anyone trying to improve how their organization decides.Chapters0:00 - Preview & Introduction1:00 - Meet James Taylor & the origins of "decision management"3:20 - Economies of scale and cross-team learning5:25 - Task automation vs. true decisioning11:35 - Decision maps and who successfully adopts them13:25 - What to do when companies say "our data is no good"18:26 - Why data science teams get undermined presenting without business input19:35 - Automating decisions with uncertainty and multiple objectives23:05 - The case for "automate first" as a default mindset28:30 - Beyond financial services: healthcare payers and personalized care plans32:45 - The Blue Polaris business story 38:15 - 3 principles for aspiring decision leadersFollow the showApple: ⁠https://podcasts.apple.com/in/podcast/the-decision-intelligence-lab/id1811085064⁠Spotify: ⁠https://open.spotify.com/show/0lFoAVKqJHTYSZNpeN61ou?si=0ae973aab0174b3b⁠Connect with JamesEmail: [email protected]: https://www.linkedin.com/in/jamestaylor/Website: https://bluepolaris.com/James's Book: https://www.amazon.de/stores/James-Taylor/author/B001IOH7UIConnect with hostsProf. Vijay Mehrotra (University of San Francisco): ⁠https://www.linkedin.com/in/vijay-mehrotra-ba9498/⁠Prof. Michael Watson (Northwestern University): ⁠https://www.linkedin.com/in/michael-watson-07600a1⁠About the podcastThe Decision Intelligence Lab podcast delivers real-world insights for data professionals, business leaders, and anyone seeking to leverage data & AI for smarter decision-making & successful business outcomes.For business inquiries, email at ⁠[email protected]⁠

Make better business decisions with data and AI—subscribe to The Decision Intelligence Lab Newsletter at ⁠⁠https://decisionintelligencelab.substack.com/⁠⁠.Meinolf Sellmann, computer scientist, entrepreneur, and founder/CEO of InsideOpt, joins the Decision Intelligence Lab podcast with hosts Mike Watson and Vijay Mehrotra. Meinolf built MIP solvers at Bell Labs, IBM, and GE before launching his startup, InsideOpt. The conversation starts with a 1998 story: George Nemhauser named three open challenges in operations research- timely decision support, multi-objective optimization, and decision-making under uncertainty. Nearly 30 years later, according to the 2025 NSF DECIDE workshop, the same problems are critical to national security and competitiveness.Meinolf explains why machine learning "hit a nerve" but couldn't deliver perfect forecasts, why optimization is seeing renewed interest, and how primal solvers attack highly combinatorial problems under uncertainty. He covers jettisoning dual bounds, scaling to 1,000 cores with ML-guided search operators, and beating Gurobi by a factor of 1,000 on quadratic assignment (Taillard instances). He shares competition wins (MaxSAT 2016, AI for TSP 2021 at IJCAI), a real coffee-roasting scheduling story, and three principles for better decision-making.What You'll Learn- Why George Nemhauser's 1998 challenges remain unsolved today.- The difference between primal solvers and dual solvers, and why dual bounds limit you.- Why perfect forecasts are impossible and what to do with residual uncertainty.- How machine learning guides search - Why a primal solver scales to 1,000 cores while MIP heuristics stall.- How Seeker beat Gurobi by 1,000x on quadratic assignment (Taillard instances).- Why the right tool beats raw algorithmic improvement.- Bridging the gap between a well-shaped technical problem and the business customer's real problem.- The coffee-roasting scheduling story — why MIP failed and a primal solver won.- Three rules for good decision-making- Why risk mitigation matters more than expected value (gambler's ruin, UPS fleet scenarios).Timestamps0:00 - Preview & Introduction0:52 - Meet Meinolf Sellmann, InsideOpt1:29 - The 1998 George Nemhauser story: Three OR challenges3:21 - Multi-objective optimization: the three canonical approaches and why they fail6:20 - Why renewed focus on decision-making after the AI/ML wave7:43 - Perfect forecasts are impossible: the sushi example9:50 - Solving combinatorially complex problems under uncertainty12:10 - What is a primal solver vs. a dual solver?14:16 - Technical problem vs. the business customer's real problem 16:45 - Jettisoning bounds; 1,000 cores; ML-guided search19:22 - Machine learning as counting cards in blackjack22:05 - Hardware vs. algorithms; beating Gurobi 1,000x on quadratic assignment23:43 - Why leave the big labs and start a company25:27 - MaxSAT 2016 win; the self-learning solver27:29 - Evolving view of good decision-making: three principles31:20 - Where to find InsideOpt and Seeker35:40 - The Coffee-Roasting Scheduling StoryFollow the showApple: ⁠https://podcasts.apple.com/in/podcast/the-decision-intelligence-lab/id1811085064⁠Spotify: ⁠https://open.spotify.com/show/0lFoAVKqJHTYSZNpeN61ou?si=0ae973aab0174b3b⁠Connect with guestMeinolf Sellmann: ⁠https://www.linkedin.com/in/meinolf-sellmann-a349636/InsideOpt: https://insideopt.com/Connect with hostsProf. Vijay Mehrotra (University of San Francisco): ⁠https://www.linkedin.com/in/vijay-mehrotra-ba9498/⁠Prof. Michael Watson (Northwestern University): ⁠https://www.linkedin.com/in/michael-watson-07600a1⁠About the podcastThe Decision Intelligence Lab podcast delivers real-world insights for data professionals, business leaders, and anyone seeking to leverage data & AI for smarter decision-making & successful business outcomes.For business inquiries, email at ⁠[email protected]⁠