There and Back Again – The Trials And Tribulations of Launching an AI Platform in the Bay Area.

February 18, 2025

Some here might be wondering: “What happened to Joe and shopperGPT ?” I was very active here for most of 2024, building relationships for the new business—clients, strategics, and investors. Then mysteriously, all communication stopped. Here’s the story about the ups and downs of building an AI startup in the Bay Area. I’ll chop this up into several posts and promise some cliffhangers.

Post 1: The Early Days and The First Setback
Could the early buzz and excitement translate into a viable product?

In 2016, I had what I thought was a groundbreaking idea: SiBi (See It Buy It), a consumer-facing AI-powered visual shopping marketplace designed to transform online fashion shopping. SiBi was built on the Shopify platform and made available in both the iOS and Android app stores. The marketplace brought together thousands of merchants and over 500,000 products, giving users the ability to search for items not just by text, but also by exact color and image similarity—a combination that no one had delivered at that point. As a serial entrepreneur with five startups under my belt—including an IPO and a successful M&A—I believed I was ready to revolutionize the online shopping experience.

But by 2018, SiBi had hit a wall. Despite strong early interest, I couldn’t secure the series seed funding needed to scale. At that time, the investment climate in the Bay Area was skeptical of B2C startups and still unconvinced of AI’s true potential. Even with my track record, raising the necessary capital proved impossible. SiBi and its parent company, nFlate, inc., ultimately shut down. I was exhausted, frustrated, and—frankly—ready to step away from the startup grind. When COVID-19 hit, I effectively retired, assuming that chapter of my life was closed.

What I didn’t know was that this wasn’t the end of the story. Far from it.

Post 2: A Spark Reignited and a New Approach
Could a shift in focus from B2C to B2B provide the breakthrough needed?

Then came ChatGPT’s launch. The buzz around this AI marvel led friends and colleagues to urge me to revisit my original concept. At first, I dismissed the idea—no way I was going back to B2C. But the growing enthusiasm around AI got me thinking: what if SiBi could become a B2B solution instead? Something that would help e-commerce companies transform their shopping experiences rather than cater directly to consumers.

At this point, major players like Google, Amazon, and eBay were leading the charge in AI search for e-commerce. Their platforms allowed consumers to instantly compare thousands of competitive products side-by-side. For many businesses, this meant a loss of control over the customer experience and a higher risk of losing shoppers to cheaper or more convenient options. Only a select few companies could afford to invest in their own cutting-edge AI search systems to keep customers on their sites. The rest were left with limited choices.

That’s when the concept of shopperGPT started to take shape. It wasn’t just a revolutionary new “search bar” concept; it was a whole new level of shopping experience. A tool that could not only understand natural language queries but also find products by color, match them visually, and provide hyper-personalized recommendations. And most importantly, it was designed to be accessible to businesses of all sizes—not just the industry giants. For the first time in years, I felt the fire of a new opportunity.

But what happens when promising ideas meet the challenges of execution?

Post 3: False Starts, Fading Partnerships, and Funding Frustration
With resources running low, could a pivot or new partnership save the day?

In mid-2023, I joined forces with a high-profile business associate who convinced me she’d be the perfect partner to deliver the re-imagined SiBi. We spent months refining the vision, but the partnership simply didn’t work out. I felt like I’d wasted precious time. In January 2024, I met someone who seemed like an ideal CTO. He brought on another engineer, and together we started building.

I dove into the front-end development headfirst, choosing React despite having no prior experience with the framework. Through trial, error, and determination, I crafted a functional front-end for the platform. Meanwhile, the engineers focused on building the backend APIs and the core AI code. Together, we had the beginnings of a working MVP—a tangible demonstration of shopperGPT ’s potential.

But six months in, the team grew restless without funding. They had full-time jobs and personal commitments (as did I). In July, the CTO told me they needed to step back until we secured financing. At the same time, I struggled to find investors who were ready to commit. It felt like every day I saw another startup announcing a funding round, and here I was, pitching relentlessly but with nothing to show for it.

I also sought guidance from two reasonably high-profile strategic advisors. Both were well-connected and genuinely interested in helping shopperGPT succeed, and both made attempts to get investors on board. But in the end, neither was able to move the needle. I had invested time and trust into these relationships, and while I appreciated their efforts, it was frustrating to realize that even their expertise wasn’t enough to push through the funding barriers.

It was maddening. I had interested parties—potential clients, strategic partners, and yes, even investors. People were clearly intrigued by the vision. But for reasons I couldn’t quite pin down, no one was writing checks. I found myself wondering: Is it the story I’m telling? Is it me? Is it the AI-for-fashion angle? The website? Maybe it’s because I was showcasing a demo app and not a fully functional store. Was it just the competition for investment in the Bay Area? I couldn’t help but think about the other startups in the region, many with impressive teams or established networks, vying for the same pool of funding. Maybe it was just the reality of trying to stand out in such a crowded field.

By December, my team had fully stepped away, leaving me with a nearly complete platform and dwindling resources. The MVP was there and the Shopify integration was well underway. We just needed to port the backend to onboard clients and build the pipeline. But my funding options were beginning to dry up just as I needed them most.

Post 4: Taking the Reins
With new determination, could a single person turn the tide?

For weeks, I wrestled with how to move forward. I considered hiring contractors, but the complexity of the task felt insurmountable. Finally, I made a choice: I would tackle the backend development myself. Even though I wasn’t a backend expert, I knew the architecture inside and out. What I lacked in experience, I’d try to make up for with persistence and the best resource I had—ChatGPT!

At first, it felt like a bold new partnership. Just me and my AI assistant. I’d say things like “Okay, ChatGPT, show me how to implement vector embeddings,” and it would respond with a helpful snippet. But, let’s be honest, it was a bit of a love-hate relationship. I’d ask a straightforward question, and ChatGPT would hand me some wonderfully detailed code that worked like a charm. Then I’d ask a follow-up and end up with a response that made me wonder if I was talking to ChatGPT or its evil twin. Sometimes it would give me code that solved one problem while creating three new ones.

Still, with every exchange—some exasperating, many productive—I slowly built the pipeline, created embeddings, and integrated the various APIs. It was actually probably like giving birth—painful, but when the first API worked, then the second—euphoria! I finally felt like I was making real progress.

I’ve written or rewritten thousands of lines of code at this stage. It’s still a little bewildering even to me. The world has changed. Readers here should take a step back to understand what this means for the future of software development. A former chemical engineer with a background in sales, marketing, and business development can learn to develop complex AI code, along with an automated client onboarding pipeline to build and populate multiple Postgres tables, in 4-6 weeks—using a combination of Google Cloud, AWS, and an open source LLM. If I had done this in July, I would have been in the market long ago, likely with comfortable funding.

Post 5: A Realization and Renewed Optimism
Can a project born out of repeated setbacks find its way to success?

As I look back, what stands out is the sheer resilience this project demanded. I’ve gone from a failed consumer startup, through a global pandemic, and finally pivoting to a B2B SaaS platform. I’ve endured failed partnerships, a lack of funding, and a team that had to step away. And yet, I’ve continued to try to find a way forward—one step at a time.

shopperGPT isn’t just about AI or search functionality. It’s about building something that fundamentally changes the way businesses and shoppers interact. It’s about creating something valuable out of setbacks and disappointments. And it’s about proving to myself that even after retirement, even after setbacks, there’s still more I can contribute. While the platform is still a few weeks away, I really feel like I can see the light at the end of the tunnel. Just hoping it’s not another freight train!

What about the “CTO factor”?
Will investors ever accept a startup without a traditional CTO?

One question I’ve been wrestling with is how investors will look at startups that no longer have a traditional “CTO” role. Historically, at least in the Bay Area, a startup with a CTO and an idea or prototype was considered fundable. The premise was that you could always find a CEO to drive the business forward if you had a solid technical co-founder building the product. But now, with the advent of AI code generation, the game is changing.

Take “Woz,” a new Y Combinator startup that’s gaining traction. Its tagline? “Your AI Technical Cofounder.” Woz promises to help entrepreneurs launch software businesses from scratch, no technical co-founder required. It’s a product of its time, emerging in a landscape where AI code assistants can accelerate development cycles and eliminate some of the barriers that once made a dedicated CTO indispensable. However, while Woz might produce usable code, the non-technical side of the business won’t have true technical ownership or a deep knowledgebase of the code it’s delivering. Without a human technical co-founder, the team may find it difficult to understand the foundational architecture, troubleshoot nuanced issues, or make informed decisions about long-term technical strategy.

The question now isn’t just “Is the product viable?” It’s also “Is the team structure still relevant in this new world of AI-driven development?” I don’t have the answers yet, but I suspect we’ll see a lot more startups challenging the traditional assumptions about what it takes to get funded. In a way, it’s an exciting time. The rules are being rewritten, and maybe—just maybe—this shift will open doors for founders who might not have had the “right” team composition in the past.

Important consideration, since I am now effectively the CEO & CTO!