Open the App Store or Google Play and you’ll find dozens of plant identification apps. Each has its strengths, but they fall short of delivering the complete experience plant enthusiasts actually need. For example, certain apps excel at community-driven identification, but get lost in scientific features when you just want quick answers. Other paid apps deliver better UX but monetize through questionable data practices. Each app optimizes for a different use case, forcing users to juggle multiple tools. But here’s the thing: plant lovers don’t compartmentalize their curiosity. The hiker wondering about trail-side wildflowers is the same person nurturing succulents at home, concerned about pet safety, and seeking care guidance.

Botanify exists to be the unified intelligence layer between humans and the plant world. We’re not building another plant ID app—we’re creating a comprehensive pocket botanist that understands context. Whether you’re hiking with your dog and need instant toxicity information, or you’re seeking personalized care guidance for your houseplants, Botanify adapts to your specific needs in that moment.

The most critical architectural decision we made was to prioritize accuracy over on-device processing. From the beginning, we recognized that plant identification models with high accuracy are too large for practical mobile distribution, so we decided to host the model on Google Cloud Run for scalable inference. This decision has proved to have positive downstream effects, as I can now update and train the identification model without being overly concerned of model size.

After using top of the line AI-assisted development tools for the past couple of months, I learnt that the biggest development constraint isn’t coding capacity, but rather, decision-making clarity. Every architectural choice, every UX decision, every feature prioritization stems from having a vision of what I’m building and why. AI can help you write code faster, but it can’t tell you which problems are worth solving. It can optimize your algorithms, but it can’t design experiences that users will love. Product vision, design intuition, and user empathy remain fundamentally human capabilities. This is what separates products that succeed from those that simply ship features. Having a clear product vision means knowing when to say no, understanding which technical trade-offs matter, and designing for real user needs rather than imagined ones.

As we scale from our current AMD 5700XT setup to more powerful hardware (hello, RTX 3060TI), we’re not just chasing higher accuracy numbers. We’re expanding our capability to handle edge cases, rare species, and regional variations that larger, less focused teams often ignore. The technical moonshots we’re pursuing:

Expanding plant coverage from ~5,000 to tens of thousands of species

Increasing accuracy of plant identifications from ~72% to ~90% for Top-1 accuracy

Insightful plant knowledge that offers users a personalized experience when serving up information accompany plant identifications

The ultimate measure of our success isn’t download numbers or retention rates—it’s whether we’ve genuinely made plant knowledge more accessible. Every accurate identification, every helpful care tip, and every prevented pet toxicity incident represents someone forming a deeper connection with the natural world. In a fragmented market dominated by feature-heavy incumbents, we believe clear vision is our competitive advantage. While others chase engagement metrics and subscription revenue, we’re solving the fundamental problem: bridging the gap between human curiosity and botanical knowledge. The bottom line: Great technical products aren’t built by accident or assembled from feature checklists. They emerge from teams that understand their users deeply, see opportunities others miss, and have the technical expertise to execute on ambitious visions.