Coding along with Gemini
Source: Dev.to
This is a submission for the Built with Google Gemini: Writing Challenge
What I Built with Google Gemini
I built a full‑stack AI‑powered pharmacy management system that helps pharmacies prevent losses from expired medicines and stock shortages. The idea came from a real conversation with a pharmacy employee who told me she was personally responsible for any medicines that expired without being reported in time. That problem needed a solution.
Working alongside Gemini as a coding partner, my process started with an idea — then I asked the AI what the best programming language and stack would be for this specific project. After evaluating the suggestions, I made the final decision myself: FastAPI for the backend, React with TypeScript for the frontend, and Groq’s free AI API for the intelligent features.
Gemini helped me write the code, but the process was far from automatic. I encountered indentation errors, broken endpoints, deprecated model names, and missing imports. I debugged using tools like Chrome DevTools, Postman, and careful error reading. The first design didn’t match my vision, so I pushed for a complete redesign. Every broken point had to be identified, understood, and fixed.
What I learned is that in this era of AI tools, the most valuable skills a software engineer can have are not just technical knowledge — they are the ability to generate real ideas from real observations, and the ability to resolve conflicts by combining tools like Gemini, Postman, and browser debugging with their own programming knowledge and creative thinking.
AI can write code. It cannot walk into a pharmacy, have a conversation, and turn that moment into a solution. That part was mine.
Demo
What I Learned
Technical skills
- Built REST APIs with FastAPI.
- Created interactive UIs with React and TypeScript.
- Managed data using SQLite and SQLAlchemy.
- Secured the application with JWT authentication and role‑based access.
- Integrated a third‑party AI API (Groq).
Soft skills
- Learned to treat AI as a tool rather than a shortcut.
- Gained confidence by reading error messages, diagnosing issues, and deciding how to fix them.
- Discovered that persistence matters more than perfection.
Unexpected lesson
The most important takeaway is that ideas are still irreplaceable. I walked into a pharmacy, had a real conversation, identified a real problem, and built a solution for it. No AI prompted me to do that. In a world where AI can generate code in seconds, the engineers who will stand out are the ones who bring the problems worth solving.
Google Gemini Feedback
What worked well
- Helpful for scaffolding code quickly, suggesting tech stacks, explaining concepts, and debugging when given the right context.
- Speed of moving from idea to working prototype was impressive.
- Valuable for learners because Gemini could explain why something works, not just what to write.
Where I needed more support
- Occasionally suggested deprecated libraries or model names, requiring extra research.
- Generated code sometimes had structural issues like incorrect indentation or misplaced logic.
- AI responses assumed a perfect environment, ignoring version conflicts, OS differences, or local setup quirks. These gaps required combining Gemini with tools like Chrome DevTools and Postman.
Overall
AI is an incredibly powerful coding partner when you bring the ideas, judgment, and debugging skills. That combination is where the real value lies.
#react, #Python, #full-stack