Available Taking on new SaaS & AI engagements

I build software that
scales in production.

I design, ship, and operate SaaS platforms and AI-driven products end-to-end — from the first commit to the SLOs that keep paying customers happy at scale.

Get in touch → How I work
~/andreaspiculell — zsh

Stack

typescript
python
go
postgres
react / next

Cloud

aws
terraform
kubernetes
~ $ whoami
andreas.piculell — software engineer · saas & ai

~ $ cat principles.json
{
"ships": "production-grade by default",
"scales": "from prototype to thousands of tenants",
"observability": "logs, traces, slos from day one",
"focus": ["saas", "ai products", "developer tooling"]
}

~ $ ./start-project --client=you
// ready when you are.
How I work

Engineering that holds up under real load.

{ }

Built for quality

Strict types, tested code paths, and CI that fails loudly. The software I ship is something you can hand to the next engineer without an apology.

∞

Designed to scale

From MVP to production traffic — multi-tenant data models, async pipelines, autoscaling, and infra-as-code that grows with the business.

⌘

Operated end-to-end

I don't hand-off and disappear. Dashboards, alerting, on-call runbooks, and post-mortems are part of every build I deliver.

What I build

SaaS platforms. AI products. The boring infrastructure that makes them reliable.

SaaS platforms b2b

Multi-tenant products that handle real customers, real billing, and real compliance — built to grow without rewrites.

  • Multi-tenant architecture with strict data isolation
  • Authentication, SSO, SCIM, and role-based access
  • Billing & subscription with Stripe
  • Background jobs, queues, and scheduled workflows
  • Admin tooling, audit logs, and customer support flows

AI products llm

LLM-powered features and agents that survive the jump from demo to paying customers — with the eval and ops to back them up.

  • RAG pipelines with evaluation and observability
  • Tool-using agents with safe, auditable execution
  • Prompt caching, batching, and cost-controlled inference
  • Anthropic Claude, OpenAI, and self-hosted models
  • Production guardrails: monitoring, fallbacks, replay
10+
Years shipping
99.9%
Uptime target
∞
Lines deleted
0
Vapourware

Have something hard
that needs to ship?

Whether you're scaling a SaaS, exploring an AI feature, or rescuing a stalled project — let's talk.

[email protected] →