Starting from Scratch

It was the beginning of May 2026, and for the first time in a long while, my life felt somewhat stable. I am currently living in a university dormitory in Germany after relocating from Ukraine. With the dust finally settling, I had a burning desire to build something meaningful with my software engineering background. I had the time, I had the drive, but I had one problem: I had no idea what to build.
I’ve always been drawn to financial markets. Recently, my focus shifted toward stocks and ETFs. I find them fascinating not just technologically, but in how their charts reflect the macroeconomic state of the world. The decision was made: I was going to build a FinTech platform.

I asked myself what tool I would actually use, and what my biggest pain point was when analyzing a portfolio. I realized there was a massive gap in the market for accessible financial backtesting tools. I didn't want an academic, terminal-style approach; I wanted a practical, visual tool for real life. So, I mapped out the architecture and prepared to dive in.

The First Attempt (And Why It Failed)

I decided to jump into the code, leveraging AI to accelerate my workflow. I rely heavily on Gemini for its benchmark-proven performance, writing out the core vision and letting it help me build.
As an engineering student, my first instinct was to design a hardcore, enterprise-grade stack:

Frontend: React and Next.js deployed on Cloudflare Pages.
Server: A 4-core, 4GB RAM EC2 instance running Debian.
Database: PostgreSQL 18 in a Docker container.
Backend: Bun.js and FastAPI. Bun’s runtime is incredibly fast and perfect for "vibe coding" with AI.
Proxy & Cache: Nginx as a reverse proxy and Valkey for caching.
IAC: Terraform scripts to manage this entire beast.

Then, I hit my first major bottleneck: AI context limits. When you are moving at the speed of thought, generating full components and routing logic, you eat through AI tokens like crazy. I was constantly hitting rate limits and context window caps. To bypass this, I pivoted my workflow, plugged the paid Google Cloud API directly into my setup, and gave the Cursor IDE a real chance. It was a game-changer for keeping the momentum going.
I had the ultimate tech stack, the AI tools, and extreme determination. I started waking up at 6:30 AM and coding until 10:30 PM. I skipped my running club workouts and ate only once a day.
Can you guess how long I lasted? Two days.
Within 48 hours, I was completely burned out. And the crazy part is that I had accomplished almost nothing product-wise. I bought domains, configured Cloudflare, set up a WAF, spun up the DB schema, and built basic auth. The rest of my time was entirely consumed by pipeline nightmares, Nginx proxy conflicts, and Docker routing issues. I had to step away.

The Vision and the Solutions I Found

While recovering, I asked myself why this was so painful. The answer was obvious: Infrastructure. I was spending 80% of my time setting up DevOps and only 20% actually writing product code.
If I wanted to move fast as a solo founder, I had to apply the golden rule: Keep It Simple, Stupid. I scrapped the complex EC2 and Terraform setup and embraced the serverless era:

Frontend: Vercel for zero-config deployments.
Backend Hosting: Google Cloud Run. It takes my container, auto-builds, and scales instances from zero to a hundred based on traffic.
Database: Neon, a serverless Postgres that handles scaling out of the box.
Framework: Bun and Elysia for blazing-fast logic.
Third-party APIs: Resend for emails, Formspree for feedback, and Vercel Analytics for tracking.

This time, the tools handled the infrastructure, giving me the mental space to actually design and build the product. You don't need to reinvent the wheel to start driving.
I also changed my schedule. I stopped the 16-hour coding binges and committed to just 4 hours of highly intense, focused work per day. Within a week, my MVP was alive. It has rough edges, but it's real, and it works.

The Takeaway

It’s not about who works harder; it’s about who works smarter and prepares for the marathon. You can have high-frequency bursts of productivity, but you must protect your rhythm, your physical health, and your personal life. Creating products on the internet is more accessible now than ever, especially if you leverage AI and managed services correctly.
I built this platform to scratch my own itch, but I believe it can help others build their edge in the market. I would be thrilled if you took 5 minutes to try it out and share your honest feedback. It means the world to a solo dev.
Keep building.
Danylo Yen

Starkon | Neo Analytics Platform

Compare crypto, stocks, and ETFs in seconds. Build your track record and share your edge.

starkon.tech