From Bootstrapped Amazon Seller to AI Studio Founder — LiliDi Blog
How Tetiana Stepanets went from bootstrapped Amazon and DTC businesses to building Lilidi.ai, a multi-model generative AI creative platform.
By Tetiana Stepanets
Before I was building an AI platform, I was selling physical products online. No venture capital. No large team. No safety net. I learned business through inventory, Amazon PPC, margins, suppliers, customer support, and the uncomfortable reality that every bad decision eventually appears somewhere in your P&L. Years later, those lessons became surprisingly relevant when I started building Lilidi.ai. From physical products to generative AI The path was not a straight line — but in hindsight, every stage taught something the next one needed. DTC & e commerce. Learning how customers actually buy, not how brands imagine they do. Amazon operator. PPC, reviews, unit economics, cash flow — the unforgiving school of margin. Product & brand building. Positioning, packaging, creative, retention. AI & SEO software. The first shift from atoms to bits: leverage, iteration, code as inventory.
Lilidi.ai. A multi model generative AI creative platform for image, video and audio. Lilidi did not come from nowhere. It is the current step of a longer operator journey. The business education Amazon gave me Selling on Amazon and running a DTC brand teaches you things that no pitch deck ever will. Inventory has a real cost. Every SKU you order is capital sitting in a warehouse until a customer decides otherwise. Acquisition costs matter. If you don't know your CAC, your ad platform will happily discover it for you at your expense. Amazon PPC can destroy your margins in a week if you don't watch it. Bad keywords bleed cash silently. Creatives directly influence conversion. A better product photo can outperform a "growth hack". Customer support is a product feature. Every ticket is a bug report about your onboarding, packaging or copy. Cash flow matters more than revenue. Growth without
cash is just a slower way to run out. Unit economics matter. If it doesn't work at 100 orders, it definitely won't work at 10,000. Operational mistakes cannot be hidden. There is no vanity metric that fixes a bad supplier or a broken SKU. The central point: bootstrapping forces clarity . When your own money is paying for inventory, software, advertising and mistakes, you learn very quickly to distinguish "interesting" from "valuable". Why I moved from physical products to software Physical products taught me discipline. Software taught me leverage. Running a physical business means dealing with manufacturing, inventory, shipping, warehousing, returns, logistics and working capital — a chain where every link can break. Software and AI have different constraints, and different upside. The reason I moved was not that "AI was trending". The interesting part was leverage: One product
improvement can immediately improve the experience for every user. One integration can unlock an entirely new creative capability. One new model can dramatically expand what creators can make. For an operator used to counting pallets, that kind of compound leverage is hard to unsee. Generative AI had a different problem When I started using creative AI tools seriously, I noticed something strange. The individual models were becoming extremely powerful. Image models, video models, voice models, music models — all improving on a monthly cadence. But the workflow for creators was getting worse, not better. One tool for images. Another for video. Another for voice. Another for music. Another for editing or upscaling. Each platform introduced another subscription, another credit system, another interface, another asset library, another way of doing things. The models were improving. The
creator's day was becoming more fragmented. That contradiction became one of the ideas behind Lilidi. What Lilidi looks like in practice Instead of explaining another AI workflow in theory, here's the product in under a minute. [[youtube:https://www.youtube.com/watch?v=INMHOQ2HYE8]] Lilidi.ai brings leading generative AI models for image, video and audio creation into one workspace. Instead of managing multiple subscriptions and disconnected workflows, creators move between different models through a unified interface and a single credit system. What is Lilidi.ai? Lilidi.ai is a multi model generative AI creative platform that gives creators and brands access to leading image, video, audio and creative AI models through one workspace. In practice, that means: One prompt bar. Multiple models across image, video and audio. A unified credit system. Transparent generation costs before you
press generate. A shared library for everything you create. Room to choose the right model for the job instead of the one your subscription happens to include. You can explore the model catalog, compare providers, or start from the pricing page. The five month sprint The current version of Lilidi was built in roughly five months by a very small team. Putting multiple AI models behind one interface sounds simple. Making them behave like one coherent product is not. Every model integration meant thinking about: Prompt structure and parameters. Generation queues and retries. Billing and credit accounting. Cost management and margin per generation. Asset handling across image, video and audio. Creative UX that stays the same no matter which model runs underneath. Testing, monitoring and support when a provider changes something upstream. A small bootstrapped team cannot afford to build all
of that with theater. You ship, you watch what happens, you fix what's broken, you keep going. What bootstrapping teaches you 1. Ship in weeks, not quarters Small teams cannot afford months of internal debate. Shipping creates information — nothing else does. The plan you had before real users is almost always wrong in ways you can't predict from a whiteboard. 2. Unit economics are product decisions Costs are not something finance figures out later. In generative AI, every generation has an infrastructure and model cost. Pricing and product architecture are the same conversation. If you don't design for margin, you're designing for a subsidy. 3. Margin is a feature Every model or feature deserves a real cost of goods analysis before it hits the prompt bar. Popular is not the same as sustainable. A trendy model that loses money on every generation is a liability disguised as growth. 4.
Founders should stay close to support I still answer support. It's the cheapest research in the world. Customer questions reveal exactly where onboarding, UX, pricing or the product itself is unclear — usually faster than any survey or dashboard. 5. Constraints create focus A bootstrapped company cannot build everything. That forces harder questions on every feature: Does this actually improve the workflow? Does it solve a real creator problem? Will people use it? Is it sustainable? Constraints aren't the enemy of a small team. They're the thing that keeps a small team from behaving like a bad big one. "Bootstrapping changes the questions you ask. Instead of 'Can we build this?' you start asking 'Does this actually make the product better?'" — Tetiana Stepanets, Founder & CEO, Lilidi.ai What changed — and what didn't On the surface, selling physical products and building an AI platform
look like completely different careers. Underneath, most of the fundamentals are identical. Understand the customer. Watch your costs. Ship. Listen. Fix what is broken. Do not confuse complexity with value. The technology changes. The operator mindset does not. What Amazon taught me about margin, inventory and support translates almost line for line into how Lilidi thinks about generations, credits and creators. Building Lilidi without the usual startup playbook Lilidi has been built with a strong emphasis on product, speed, users and sustainable economics. Not against venture capital — just without needing it to make the next decision. A small team has real advantages: Short feedback loops between creators and code. Fewer layers between an idea and a shipped feature. Faster decisions when a provider changes their pricing or a model breaks. Direct founder involvement in the parts of the
product that matter most. Rapid iteration on the things users actually complain about. The tradeoff is honest: a small team has to prioritize aggressively. Not every good idea gets built. Not every trend gets chased. Every "yes" is also a "not yet" to something else. That's fine. It's the same tradeoff every operator has to make, at every stage. For the deeper story behind why I started the company, read Why I'm Building Lilidi AI or visit the founder page. FAQ Who is Tetiana Stepanets Tetiana Stepanets is an entrepreneur and the Founder & CEO of Lilidi.ai. Her background spans DTC e commerce, Amazon selling, product development and AI software. She has spent most of her career as a bootstrapped operator building products for real customers rather than as a career executive. What did Tetiana Stepanets do before Lilidi.ai Before Lilidi.ai, Tetiana built and operated businesses across e
commerce, DTC and product development, including a physical goods brand sold on Amazon. That operator experience — margins, inventory, PPC, support — shaped how she now approaches building an AI platform. Was Tetiana Stepanets an Amazon seller Yes. Tetiana operated a physical goods brand on Amazon before moving into AI software. The Amazon experience — PPC, unit economics, cash flow and customer support — is one of the reasons Lilidi is built around margin, credits and transparent generation costs rather than pure growth at any cost. Why did Tetiana Stepanets move from e commerce to AI The move was about leverage. Physical products scale linearly with inventory, shipping and logistics. Software and generative AI scale differently — one improvement can benefit every user, and one integration can unlock an entirely new creative capability. That combination made the shift feel logical
rather than trendy. What is Lilidi.ai Lilidi.ai is a multi model generative AI creative platform. It gives creators and brands access to leading image, video, audio and creative AI models through a single workspace, with one prompt bar, one credit system and one shared library instead of a stack of separate subscriptions. Is Lilidi.ai a bootstrapped AI startup Yes. Lilidi is built by a small team with a strong emphasis on product, speed, users and sustainable economics. The current product was shipped in roughly five months without outside capital, and pricing, credits and margins are treated as first class product decisions. What does Lilidi.ai do Lilidi.ai lets creators generate and edit images, video and audio using multiple leading AI models from one interface. Instead of switching between separate tools, users pick the right model for the job — image generation, video generation,