What I Learned as a Startup CTO in Singapore

From late 2020 to early 2022, I served as the CTO of an early-stage startup in Singapore. It was one of the most challenging and rewarding experiences of my career. Here is what I learned.

The Context

The startup was building a content recommendation platform. When I joined, we had:

• 3 people (including me)
• A vague product vision
• 6 months of runway
• Zero technical infrastructure

By the time I left, we had:

• A working product with real users
• A recommendation engine with 85% accuracy
• A small but effective engineering team
• Extended runway through a seed round

Technical Decisions That Mattered

1. Recommendation Engine Architecture

Building a recommendation engine from scratch was our core technical challenge. We went with a hybrid approach:

Collaborative Filtering (LightGCN)

• Graph-based neural network for user-item interactions
• Better than traditional matrix factorization for sparse data
• Required significant GPU resources for training

Content-Based Fallback (ALS)

• Solved the cold-start problem for new users
• Used item features when we lacked interaction data
• Much cheaper to run

The key insight: you need both. Pure collaborative filtering fails for new users, and pure content-based misses the serendipity that makes recommendations feel magical.

2. Backend Stack Choice

We chose NestJS with GraphQL. In retrospect, this was the right call:

Why NestJS:

• TypeScript-first with great DX
• Modular architecture scales well
• Excellent decorator-based patterns

Why GraphQL:

• Mobile clients could request exactly what they needed
• Single endpoint simplified client development
• Great tooling (Apollo, code generation)

3. Infrastructure Decisions

We went all-in on AWS with:

• ECS Fargate for containers
• RDS PostgreSQL for primary data
• ElastiCache Redis for sessions and caching
• S3 + CloudFront for media

What I would change: I would seriously consider serverless-first today. Our infrastructure costs were significant for a startup, and managing containers was overhead we did not need early on.

Leadership Lessons

Hiring is Everything (and Incredibly Hard)

As CTO, I spent ~40% of my time on hiring. Key lessons:

1. Culture fit matters more than skills - You can teach technology, not values
2. Take-home assignments reveal more than interviews - We learned more from a 4-hour project than 10 hours of interviews
3. Reference checks are underrated - Actually call references, ask specific questions

Technical Debt is a Business Decision

Every shortcut has a cost. I learned to frame technical debt in business terms:

"We can ship this feature in 2 weeks with technical debt, or 4 weeks without. The debt will cost us ~1 week per month in slowdown. When do we need this feature to hit our metrics?"

This framing got much better buy-in from non-technical co-founders.

Communication is Your Primary Job

As CTO, your job is not writing code. It is:

1. Translating business needs into technical requirements
2. Communicating technical constraints to business stakeholders
3. Aligning the engineering team around shared goals
4. Unblocking people who are stuck

I probably spent 60% of my time in meetings, 20% on architecture/code review, and 20% writing code myself.

What I Would Do Differently

1. Hire a Senior Engineer Earlier

I tried to do too much myself. Bringing in a senior engineer 3 months earlier would have been worth the runway cost.

2. Set Up Better Processes from Day One

We operated in chaos mode for too long. Simple processes like:

• Weekly sprint planning
• Daily standups
• Code review requirements

These are not bureaucracy. They are communication infrastructure.

3. Take More Breaks

Startup life is intense. I burned out multiple times because I did not protect my recovery time. Sustainable pace matters more than heroic sprints.

Was It Worth It?

Absolutely. Being a startup CTO taught me:

• How to make decisions with incomplete information
• How to build and lead a team
• How to balance technical idealism with business pragmatism
• What I actually value in work

Even though the startup did not become a unicorn, the experience was invaluable. If you are considering a similar role, I would encourage you to go for it - but go in with realistic expectations.

The best career growth often happens in the most uncomfortable situations.