Skip to content
← Insights
TechnologyStrategyStartups

Choosing Your Tech Stack: A Fractional CTO's Perspective

Arshad Ali · November 14, 2026

Founders ask about tech stacks more than almost anything else. The honest answer is that the stack matters less than people think, and the wrong reasons for choosing one matter more. Here is how we approach the decision when we work as a fractional CTO, and the defaults we recommend for most new products.

The Principles Behind the Choice

  • Optimize for speed of change, not theoretical scale. Your first architecture needs to survive constant change to the product. Millions of users is a problem you'd be lucky to have, and one you can address when you get there.
  • Choose boring technology. Mature tools have documentation, answers to common problems and fewer surprises. Save your novelty budget for the thing that makes your product unique.
  • Make sure you can hire for it. A stack only works if you can find people to build on it. Check the hiring market before committing.
  • Buy what isn't your product. Authentication, payments, email and search are solved problems. Every hour spent rebuilding them is an hour not spent on your customers.
  • Minimize the number of languages. A small team running three languages carries three sets of tooling, dependencies and context switching.

Our Default Recommendations

These are starting points, not rules. Each one has exceptions, but they hold for most B2B SaaS products.

Language: TypeScript, end to end

One language across front end, back end and infrastructure code means shared types, shared validation and engineers who can work anywhere in the codebase. Python is the strong alternative when the product is centered on data science or machine learning.

Front end: React

Not because it is the best in every dimension, but because its ecosystem, component libraries and hiring pool are unmatched. Make sure public pages are rendered as real HTML if they need to rank in search.

Database: PostgreSQL by default

A relational database handles the widest range of needs, including reporting, flexible queries and transactions. Choose DynamoDB when your access patterns are well known in advance and you want hands-off scaling, as is often the case in serverless products. Avoid running both from day one unless you have a clear reason.

Compute: managed services on one cloud

Serverless functions or managed containers, rather than servers you patch yourself. On AWS that usually means Lambda with API Gateway for event-driven and spiky workloads, or ECS on Fargate for long-running services. Kubernetes is rarely worth its operational cost for an early-stage team.

Identity: use a managed provider

Amazon Cognito, Auth0 or a similar service gives you secure sign-up, multi-factor authentication and enterprise single sign-on without building it yourself. Enterprise customers will ask for SSO sooner than you expect.

Infrastructure as code from the first day

Define your infrastructure with AWS CDK or Terraform, even when it's small. It makes environments reproducible, changes reviewable and disaster recovery possible. Resources clicked together in the console are one of the most common sources of "it works in staging" problems.

AI features: call model APIs behind your own interface

Start with hosted models through Amazon Bedrock or a provider's API. Wrap them behind your own internal interface so you can switch models as prices and capabilities change, which they do every few months. Fine-tuning and self-hosting can wait until you have data showing they're needed.

What We Talk Founders Out Of

  • Microservices for an MVP. A well-structured single application is faster to build and change. Split later, along boundaries you've learned from real use.
  • Choosing a stack because a big tech company uses it. Their constraints aren't yours.
  • Multi-cloud "to avoid lock-in". For a startup, the cost of portability usually outweighs the risk it protects against.
  • Building an internal framework. If the first months go into tooling, the product waits.

Things to Get Right Regardless of Stack

  • Automated deployments, with a pipeline that runs tests on every change.
  • Separate environments for development, staging and production, ideally in separate cloud accounts.
  • Logging, error tracking and basic metrics before your first customer, not after your first outage.
  • A tenant or organization identifier in your data model from the start, if you're building B2B SaaS.
  • Secrets in a secrets manager, never in code or in environment files checked into source control.

When to Revisit

A stack decision isn't permanent, but changing it is expensive, so revisit it when there is evidence rather than enthusiasm: a specific bottleneck you can measure, a hiring problem you can't solve, or a customer requirement the current stack can't meet.

The best stack is the one your team can ship and operate confidently. If you're about to make these decisions and want a second opinion from someone who has made them many times, book a conversation with us.

Next step

Tell us what you’re working on

Send a short description of your project or problem. You’ll get a reply from our principal architect within two business days, with an honest view on whether we’re the right fit.