I'm building an interview prep platform nights and weekends. Here's everything that's gone wrong so far.
About a months ago I started building TierOnePrep — an interactive interview prep platform for system design and coding interviews — while keeping my full-time job as an SRE at a large tech company. Here's the honest version of how it's going, including the mistakes.
Why this product
I've conducted a lot of interviews over 12+ years. The same thing kept happening: strong engineers failing system design rounds, not from lack of skill, but because the interview format is a skill nobody teaches. Existing prep is either static articles or $150+ human mocks. And nothing covers what AI-era companies are now asking — LLM inference, GPU scheduling, vector databases. That gap felt worth building for.
Mistake #1: I picked a cloud region that doesn't support the feature I needed
I deployed on Cloud Run in asia-south1 (my main audience is in India, so it made sense for latency). Then I discovered that region doesn't support custom domain mappings. So my app was live, but I couldn't point my domain at it the normal way. The fix ended up being a Cloudflare Worker acting as a proxy in front of Cloud Run. It works well, but I lost a weekend to something a 5-minute region-check would have prevented. Check your region's feature support before deploying, not after.
Mistake #2: UUID URLs almost killed my SEO before it started
Every question page had a URL like /questions/8f3a2c91-4b7e... — completely meaningless to Google and humans. My SEO audit flagged it as the single biggest issue on the site: no keywords in URLs, duplicate title tags, markdown syntax leaking into meta descriptions. I'm still working through the fixes. If you're building anything content-driven: design your URL structure on day one. Migrating URLs later means redirects, re-indexing, and lost time.
Mistake #3: Almost falling for the backlink directory trap
I recently got handed a list of 40 "directory submission sites" for backlinks. A decade ago that might have worked; today it's a great way to build a spam profile Google penalizes. I threw the list away. The boring truth: a Product Hunt launch, real community posts, and content that earns links beat 40 junk directories.
What's actually working
A free tier (12 questions, no card) as the top of funnel
Interactive format as the differentiator — system design questions are 4-phase interview simulations, coding questions have live algorithm visualizations
Regional pricing so engineers in India aren't priced out ($11.99 US / ₹499 India)
Paddle over Stripe as merchant of record, so I don't handle VAT across countries myself — worth it for a solo operation
~92 pages indexed in Search Console and climbing
What I'm still figuring out
Distribution. Building the product was the easy part — I'm an engineer, that's my comfort zone. Getting engineers to find it is the actual hard problem. Currently betting on SEO + a Product Hunt launch + shipping new questions weekly.
Happy to answer anything — especially from others building content/education products while employed full-time. And if you've cracked the "engineer builds product, product needs marketing" problem, I'm all ears.
Good call locking in MoR early, rare to see a solo founder think about VAT before it becomes a fire to put out. One thing worth knowing as you scale: MoR providers vary a lot on take rate and what's actually included (chargeback handling, payout speed, support responsiveness).
I like that you're treating interviews as a skill in their own right rather than assuming technical knowledge is enough.
The line that stood out was strong engineers failing system design rounds despite being capable engineers. That's a very different problem from "people need to learn more." Helping someone demonstrate what they already know feels like a much stronger position than simply offering more interview content.
You’ve put your finger on the exact distinction, and honestly said it more crisply than I have. The market is flooded with “learn more” — more questions, more videos, more courses — but most people failing these rounds aren’t under-skilled, they’re under-rehearsed at showing the skill under interview conditions. Those are completely different problems, and they need different solutions.
An engineer who designs distributed systems at work every day can still freeze when handed a vague prompt and a blank canvas with someone watching. That’s not a knowledge gap — it’s a performance-under-a-specific-format gap. That’s the thing I’m actually trying to close.
Appreciate you articulating it this clearly — it’s genuinely helped sharpen how I think about what this is for.