Stanford University

Hands on education about AI Agents (TECH 2025)

Guest Lecturer

OSS4AI

150k+ Online Community

13k+ Newsletter Subscribers

17k+ Event Calendar Subscribers

Seven figure business

Chief Executive Officer

Banyan

Venture Partner

LinkedIn

Teaching 1000s of developers how to build AI applications

Advanced RAG: https://www.linkedin.com/learning/hands-on-ai-introduction-to-retrieval-augmented-generation-rag

Intro RAG: https://www.linkedin.com/learning/advanced-rag-applications-with-vector-databases/rag-with-vector-databases-advanced-strategies-for-ai-optimization

LinkedIn Learning Instructor

AICamp

Helping organize educational talks and networking events for the AI Community

Community Manager

PythonAlgos

Blogger

Zilliz

Senior Developer Advocate

Hiring new DevRel, managing the DevRel strategy, and leading the team. Developing and writing tutorials for proof of concepts for large language model applications such as: multi-document querying, conversation chaining, and more. Giving talks about vector databases, LLM Apps, semantic search, and tangential spaces.

Developer Advocate

MLOps Community

Grew MLOps Seattle's Community from 150 to 700+ (unfortunately) without help from the organization at large, which is why I left.

The Text API

Founder

Deepgram

Helped strategize ML Engineer Growth Strategy
Helped turn Python into one of the fastest growing segments of the user base
Wrote tutorials on PyTorch, BERT, and Deepgram

Technical Evangelist Consultant

Yujian Tang started developing software professionally at the age of 16. In college, he studied computer science, neuroscience, and statistics and published machine learning papers to conferences like IEEE Big Data. After graduation, he worked on the AutoML system at Amazon before moving on to build his own companies including a data aggregation app, an NLP API, and his current company - OSS4AI, an organization aimed at providing all developers access to the resources to understand, use, and contribute to the direction and development of AI.

Yujian Tang

Big thanks to everyone who made the SVAI Demo Day #2 a huge success last night

Special thanks to Snowflake Macroscope and Autonomy for making the event possible 

Congrats to our demo day winner - Flexprice ! Among 8 demos, our industry expert judges Victor Barros Anrui (Allie) Gu Anubhav Sharma and Devang Gaur all voted favorably for your project

Hope to see you all again soon - maybe at Seattle Startup Summit ;) or even earlier

Guest Lecturer @ Stanford University | CEO @ OSS4AI

Bit Bros LLC

💹 AI-Powered, Human-Centered Decisions

➤ Co-authored “You Got the Data Job… Now What?”, a career guide helping data professionals accelerate career growth and break into leadership roles.
➤ Designed career acceleration programs, guiding professionals on strategy, influence, and leadership development.
➤ Built a community-driven learning approach, helping professionals navigate career transitions and land top-tier roles.

Co-Founder | Author | Tech Strategist

LeetCode

🖥 Shaping the Future of AI & Data Science Education

➤ Orchestrated the world’s best DSA content for developers, creating structured learning paths for millions of users.
➤ Developed AI, data science, and algorithms content, increasing user engagement by 35%.
➤ Published 100+ AI/ML problems and solutions, making LeetCode a go-to resource for AI and DS professionals.

Data Content Lead

➤ Crafted high-quality Pandas and SQL content, shaping the learning experience for data professionals worldwide.

Solution Author

Toyota North America

🚀 Driving Digital Innovation with AI-First Decision Intelligence

➤ Part of a handpicked team leading Toyota’s AI transformation across North America, with a clear mandate to scale globally. Rearchitecting how a world-class manufacturer operates, using AI to deliver unmatched customer experiences.
➤ Partnering with global executives to embed decision-making excellence across domains. Architecting Toyota’s enterprise decision intelligence framework: integrating optimization, simulation, and probabilistic forecasting into global operations.
➤ Championing a culture shift across North America: from data dashboards to decision-centric systems, empowering teams to act, not just analyze.

Principal Engineer | Digital Innovations

🚘 Powering the World’s Largest Automaker with AI-First Supply Chain Decisions

➤ Led Toyota’s AI-driven supply allocation overhaul, increasing production by 40,000 units/year and delivering $390M in profit.
➤ Managed a $5M budget and a 12-person cross-functional team to build Toyota’s data-driven vehicle ordering system, optimizing inventory and adding $64M in profit.
➤ Developed an AI-powered accessories recommendation engine that optimized $700M in inventory and drove $42M in revenue.
➤ Advised senior executives on AI strategy, embedding decision intelligence into long-term global planning.

Principal Decision Scientist | Supply Chain

Simplilearn

🤖 Empowering Future AI Leaders with Caltech-backed Education

➤ Partnered with Caltech to deliver advanced AI/ML training, equipping professionals with industry-ready skills.

AI & Machine Learning Bootcamp Instructor

Infosys Consulting

📈 Enterprise AI Strategy & Automation for Fortune 50 Clients

➤ Led enterprise AI adoption for a Fortune 50 logistics firm, increasing operational efficiency by 20% and reducing costs by 30%.
➤ Defined AI roadmaps and scalable deployment strategies, ensuring long-term impact across business units.
➤ Won the Infosys Technology Gold Award for leading the successful buildout and production deployment of a global, low-latency prediction system that scales to billions of daily inference calls a day

Senior AI Consultant

Ford Motor Company

📍 AI & Optimization-Driven Mobility Solutions

➤ Developed Ford’s first revenue-generating AV subscription product, a discrete event simulation tool, collaborating with giants such as Uber, Lyft, John Hopkins University, and Walmart. 
➤ Designed fleet scheduling algorithms that cut fleet size by 27% and optimized 200,000+ trips, improving efficiency, reducing costs, and enhancing customer satisfaction.

Sr. Data/Operations Research Scientist | Autonomous Vehicles

➤ Designed, developed, and patented a point-to-point mobility system using dynamic programming to optimize for public transit.

Data/Operations Research Scientist | Smart Mobility

Marposs

💡 Automating Analytics & Software Development

➤ Engineered robotic programming solutions to enhance quality control, reducing defects and improving manufacturing efficiency.

Software Engineer

➤ Built a Python-based automation tool, reducing manual reporting time by 30%.

Data Analyst

Oakland University

Mathematics Lecturer

FCA Fiat Chrysler Automobiles

Data Mining Research Assistant | Warranty & Reliability

Auto-Owners Insurance

Software Developer

For nearly a decade, I’ve specialized in turning AI from a buzzword into a catalyst for profitable, sustainable, and future-proof enterprises. My work revolves around decision intelligence, multi-objective optimization, and enterprise AI adoption, ensuring organizations don’t just use AI but thrive because of it.

Over the years, I’ve led large-scale AI transformations within Fortune 50 environments, orchestrating complex supply allocation and inventory optimization initiatives. By combining operations research, cloud-based MLOps, and reinforcement learning, I’ve developed end-to-end AI solutions that scale globally, increase production capacity by tens of thousands of units, and generate hundreds of millions in added profit. Whether it’s building AI-powered recommendation engines that optimize billions of dollars in inventory or forecasting systems that align production with real-time market dynamics, my focus is always on driving measurable, strategic impact.

But implementing AI at scale isn’t just about models and code—it’s about influencing C-suite stakeholders, managing change, and embedding AI into an organization’s very DNA. I’ve partnered with executives and cross-functional leaders to ensure AI insights become actionable, real-time decisions that align with business objectives. That means navigating complex governance, bridging AI innovation with operations, and ensuring agile, data-driven cultures can flourish.

My expertise spans ML frameworks, cloud AI (AWS, Azure, GCP), PySpark, distributed systems, and microservices architecture—tools that empower enterprise-grade scalability. Yet, I believe tools are secondary to the vision and strategy behind them. True AI success comes from orchestrating a multidisciplinary approach, weaving together data science, operations research, and executive strategy to drive top- and bottom-line growth.

If you’re an organization looking to elevate AI into a competitive advantage—balancing profitability, sustainability, and market share—I’m ready to help you chart that course. Let’s redefine how decisions are made, seize opportunities in uncertainty, and unlock long-term competitive advantage through AI.

Send me a message or connect—let’s talk about transforming your business with AI-driven decisions.

Adam DeJans Jr.

Breaking: Local man perfectly predicts he will get wet, then jumps into pool without towel, swimsuit, or phone in waterproof case.

“But I was so accurate!” he shouts, while his iPhone dies.

This is your supply chain on forecast accuracy metrics. 😏 

Congratulations! You predicted demand would be 1,000 units. Actual demand was 1,003 units. You’re 99.7% accurate!

You also ran out of stock on day 2, airfreighted emergency inventory at $50k, and your competitor took your customers.

But hey, great forecast! 🎊

Now let’s ask the real questions… did your profits go up? Did inventory costs go down? Or did you just… forecast better?

Nobody wants to admit that forecast accuracy ends up as a vanity metric.

You can have a 95% accurate forecast and still make catastrophically bad decisions. 

You can also have a 70% accurate forecast and print money.

Why? Because the real world doesn’t care about your MAPE score.

The real world cares about:
• Did you stock out during peak season?
• Are you sitting on $2M of dead inventory?
• Did you airfreight products at 10x cost because your “accurate forecast” didn’t account for lead time variability?

Forecast accuracy measures how well you predicted the past. Decision quality measures how well you’re preparing for the future.

These are not the same thing.

A good decision framework accounts for uncertainty, asymmetric costs, and the economic impact of being wrong. It asks: “What should I do given what I don’t know?” not “Look how well I predicted this number!”

Stop optimizing for forecast accuracy. Start optimizing for decisions that make money.

Your CFO will thank you.

What’s the worst “we improved accuracy!” celebration you’ve witnessed that changed absolutely nothing?

#SupplyChainOptimization
#DontBeADumDum

Optimization @ Gurobi | Author of the MILP Handbook Series

Join me tomorrow for a deep dive into Gurobi log interpretation and model tuning!

Tomorrow, I’ll be leading a session on how to interpret Gurobi log files and use these insights to tune and improve your mathematical optimization models.

Understanding the Gurobi log can help uncover how the solver interacts with your model. This can help identify performance bottlenecks and making data-driven decisions to improve runtime and solution quality. We’ll walk through real examples, discuss key parameters, and explore strategies for efficient model tuning.

📅 Time: 12:30 PM – 1:30 PM EST
📍 Where: Whova app: https://lnkd.in/eD_MDqWp

If you haven’t signed up yet, there’s still time! This session is especially valuable for anyone looking to move beyond “it runs” to “it runs efficiently.”

💡 Bonus: Completing select sessions counts toward the Gurobi Opti301 Certificate, a great way to demonstrate your advanced understanding of optimization and solver best practices.

People often ask whether pricing can be optimized. The answer is yes... but only if you are optimizing the right thing.

It is not the prices themselves that should be optimized. It is the pricing strategy.

That may sound like a subtle distinction, but in practice, it changes everything.

Retail prices are not static decisions. They are outcomes of a complex environment. Costs shift. Competitors react. Demand fluctuates. A price that works in January may be a mistake by February. Trying to optimize a specific number in that context is like trying to hit a moving target while the wind is changing.

But pricing strategies is where we have control.

A pricing strategy is a rule. It is a logic that takes the current environment and turns it into an action. It tells you what price to post, given what you know. That is the decision. And that is what we can test, compare, and improve over time.

When we run experiments, we are not asking whether $19.99 beats $17.49. We are asking whether Strategy A, which might lean on cost-plus logic, outperforms Strategy B, which might use elasticity estimates and competitor tracking.

And we can do that experimentally. If I have 200 products, I can apply Strategy A to half and Strategy B to the other half. Let them run. Prices will change daily, even hourly. But over time, I will see which rule generates more margin, higher conversion, or better sell-through (whatever outcome I care about).

This is not about locking in a number. It is about finding the decision logic that learns and adapts with the market.

In Sequential Decision Analytics, we do not fixate on the outcome of one decision. We focus on the policy: the mapping from information to action. That is what gives us flexibility. That is what makes experimentation meaningful. And that is what allows us to learn systematically.

In pricing, as in most dynamic environments, we do not optimize answers. We optimize policies.

And that shift in mindset changes how we build, test, and improve every decision we make.

#PricingStrategy #DecisionIntelligence #SequentialDecisionAnalytics #DynamicPricing #PolicyOptimization #RetailAnalytics #ABTesting

Most decisions in the real world come with constraints. In supply chains, in finance, in energy systems, I rarely see a problem that is unconstrained. The challenge is not just choosing an action, it is choosing one that fits within limits.

Some constraints are grounded in physics and economics. You cannot ship what you do not have. You cannot store more than your warehouse can hold. A supplier might only sell in truckload increments, or impose a minimum order quantity. These are hard constraints, and they shape the feasible set of actions.

But not all constraints come from the outside world. Many come from inside the organization.

I have seen budgets locked in months before actual sales were known. A department might be given a $300M cap on annual purchasing, not because that number was optimal, but because it was available. The budget is fixed. The environment is not. And the decision-maker is now stuck navigating a constraint born of a meeting, not a model.

This is where sequential decision-making requires more than clever optimization. It requires thoughtful design.

When I build decision systems, I begin by separating real constraints from artificial ones. The first category is unavoidable. The second can be challenged. A budget constraint may make sense when decisions are made manually and infrequently. But if the system can re-evaluate priorities daily, using current data, then a static budget becomes an unnecessary straightjacket.

Good decision design means surfacing the true drivers of value. That includes understanding why a constraint exists, what risk it is meant to manage, and whether it still serves that purpose. Often, I find that constraints were introduced to simplify a broken process. Once the process improves, the constraint can be removed.

Sequential Decision Analytics gives us a way to test these ideas in practice. We can simulate trade-offs, evaluate new policies, and determine whether the business is better off with a flexible rule or a fixed one. We stop relying on inherited rules and start learning from experience.

The goal is not just to make better decisions within constraints but also improve the constraints themselves when possible.

We do not have to accept every limit as permanent. Some are just placeholders for a better system we have not yet built. And when that system arrives, we owe it to the organization to set it free.

As a technical member at Gurobi, one of the biggest “aha” moments I try to share with customers is that real numbers aren’t “real” in software. We all run on IEEE-754 double precision, where 1 == 1 + 1𝘦-16 can evaluate to true and even addition isn’t strictly associative. This is the terrain your optimizer is driving on. 

Why this matters for models:

📌 If you round inputs (e.g., write 0.333 instead of the exact fraction or more precise decimal), you can change presolve algebra and end up with a different (and wrong) elimination path. Keep full precision or clear denominators algebraically.

📌 Choose sane scales. Defaults like 𝘍𝘦𝘢𝘴𝘪𝘣𝘪𝘭𝘪𝘵𝘺𝘛𝘰𝘭 = 1𝘦-6 and 𝘐𝘯𝘵𝘍𝘦𝘢𝘴𝘛𝘰𝘭 = 1𝘦-5 work well when coefficients, RHS, and variable ranges live in human territory. Extremely tiny or huge ranges amplify floating-point limits and make “almost parallel” rows toxic. Rescale variables/units when needed.

📌 Be careful with Big-M. Coupling a large M with integrality tolerance can admit tiny “pseudo-on” binaries that still pass feasibility, while letting continuous companions drift. Prefer indicator constraints or SOS where possible. 

📌 Use symbolic infinity correctly. In the APIs, use 𝘎𝘙𝘉.𝘐𝘕𝘍𝘐𝘕𝘐𝘛𝘠 (≈ 1𝘦100); anything above ~1𝘦30 is treated as infinite for bounds anyway. Don’t invent your own mega-constants. 

📌 When numerics bite and you can’t reformulate further, try surgical parameters: 𝘕𝘶𝘮𝘦𝘳𝘪𝘤𝘍𝘰𝘤𝘶𝘴, 𝘚𝘤𝘢𝘭𝘦𝘍𝘭𝘢𝘨, 𝘖𝘣𝘫𝘚𝘤𝘢𝘭𝘦; test presolve variants (𝘈𝘨𝘨𝘳𝘦𝘨𝘢𝘵𝘦=0, 𝘗𝘳𝘦𝘴𝘰𝘭𝘷𝘦=0) and, if barrier misbehaves, try simplex or concurrent. Diagnose; don’t blindly tighten tolerances.

Numerics are part of the modeling craft. If you respect the geometry and the scales, Gurobi will reward you with speed and reliability. If you fight the floating-point world, it quietly fights back.

taplio

The Battle

Yujian Tang

Adam DeJans Jr.

Followers

Posts

Engagements

Likes

Performance ScoreYujian Tang's LinkedIn Performance

Performance ScoreAdam DeJans Jr.'s LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Yujian Tang

Adam DeJans Jr.

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Yujian Tang's LinkedIn Performance

Performance ScoreAdam DeJans Jr.'s LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreYujian Tang's LinkedIn Performance