Hands-on CTO · Product architect · AI engineering

Build the product.
Grow the team.
Make delivery reliable.

I turn ambitious ideas and complex domains into working products, capable engineering teams and reliable delivery systems. I bring hands-on technical leadership to startups and established companies introducing AI.

Permanent leadership roles · AI engineering consulting · Fractional CTO

Building software since 2008 CTO and technical leadership Founder and operator

Discuss a role or project See selected work
PRODUCT / SYSTEM / TEAM
Dan Avramescu, hands-on CTO and product architect
Dan Avramescu Strategy connected to delivery
GLOBAL / REMOTE
TOP 3% TALENT Vetted by
Since 2008Building software
Zero to productArchitecture and delivery
Team formationRoles, practices and ownership
Founder/operatorCommercial and operational reality

Selected work

Products built. Teams formed.

From healthcare procurement to business-facing AI: product definition, hands-on architecture and the engineering teams and practices behind delivery.

Grapevine medical procurement product interface 01 / PRODUCT + TEAM

Grapevine Technologies

From an idea to a working product and a 12-person engineering organization.

10 monthsStarting in 2020: product, architecture and team formation.

Clinics buying medical supplies often relied on a small network of contacts. Finding another source meant more calls, limited price visibility and additional layers of intermediaries.

I helped turn that domain knowledge into Grapevine’s first customer-facing product and built the engineering organization from zero to 12 people.

  • Translated procurement expertise into product requirements, search and data architecture.
  • Built the initial team and established responsibilities and delivery practices.
  • Helped deliver ordering across a wider range of suppliers, including importers closer to the source, giving customers access to pricing beyond their existing supplier network.

The product later expanded to support larger customers and more complex supplier and volume-based commercial terms.

What this provesDomain translation · Product delivery · Engineering team formation
Visit Grapevine (opens in a new tab)

OptAI: Lotti and Huusund

AI that fits into the business, with less work for the people using it.

Across four product pivots at OptAI, the central question has become clearer: how much useful work can software do without asking a business owner to manage another tool?

I lead product and technical architecture and work hands-on on the systems. The latest two directions, Lotti and Huusund, bring that thinking to AI-assisted social media with minimal onboarding and ongoing effort.

  • Conversational workflows that connect business context, content creation, review and publishing.
  • Human control at important decisions, with explicit approval boundaries.
  • Multi-tenant architecture, background processing and recovery when external services fail or return uncertain results.
Design focusLow-effort adoption · Human control · Reliable AI operations
Explore Lotti (opens in a new tab) Explore Huusund (opens in a new tab)
Lotti AI-assisted social content product 02 / AI + OPERATIONS

Where I am most useful

Bring me the difficult transition.

You do not need to arrive with technical language. Start with the business situation that is not moving.

“We have the domain knowledge but do not know how to turn it into software.”
“Our prototype works, but it is not yet a reliable product.”
“We need a technical team and someone capable of building it.”
“Our AI workflow works in a demo but not as an operating business system.”
“We have accumulated technical complexity and need a clear path forward.”
“We need senior judgment and someone who can work directly on the product.”

How I help

Leadership first. Hands-on throughout.

I am prioritizing permanent leadership roles, followed by AI engineering consulting and fractional CTO engagements.

Agentic engineering

Faster implementation needs stronger verification.

I use Claude and Codex inside a delivery process built around scoped work, independent review and evidence. My responsibility is still the product, the decisions and acceptance of the result.

Read how I build and verify with AI agents
  1. 01

    Define the work

    Scoped tickets, acceptance criteria and a second opinion on the plan before implementation.

  2. 02

    Give agents bounded ownership

    Isolated worktrees and controlled permissions let work progress with clear boundaries.

  3. 03

    Test the actual journey

    Independent review, evaluations and conversation tests exercise real services and persisted state. That test lab uncovered a production routing defect missed by unit tests.

  4. 04

    Make verification efficient

    Better CI hardware, optimized workflows and conditional test suites reduce unnecessary work. Subsequent commits should not keep obsolete runs consuming resources.

  5. 05

    Keep human acceptance

    Passing checks is evidence for a decision. I remain responsible for whether the change meets the product’s needs.

About Dan

A builder who moves between product, software, teams and operations.

I have built software since 2008 and spent much of that time where product uncertainty and technical responsibility meet. My work repeatedly starts with a difficult question: what should this product actually do, and what system and team can make it real?

My main work spans healthcare procurement, business-facing AI products and engineering team formation. Outside that, Fable Steps is a personal storytelling project. I also co-created Hubuddha Villas in Ubud several years ago; the business now operates independently of me.

I care about adoption as much as architecture. A product succeeds when it fits naturally into how people already work, gives them appropriate control and can be operated reliably by the team behind it.

Technical range

Enough depth to connect the whole product.

Technical breadth matters because product risks rarely stay inside one layer.

Product systems
  • SaaS foundations
  • Authentication, organizations and permissions
  • Payments and subscriptions
AI and data
  • AI workflows, evaluations and guardrails
  • Human review and approval
  • Search, data architecture and integrations
Delivery systems
  • APIs and background processing
  • Cloud deployment and observability
  • Conditional CI/CD, retries and reconciliation
Technical leadership
  • Product architecture and roadmaps
  • Engineering team formation
  • Development and quality practices

Working toolkit: TypeScript, JavaScript, React, Next.js, Node.js, Python, PostgreSQL, MySQL, OpenAI, Claude, Docker, cloud deployment and CI/CD.

Working fit

The right conditions for useful work.

Strong fit

  • Startups building a product and engineering organization
  • Direct access to decision-makers
  • Need for strategy and implementation from the same person
  • Established companies introducing AI into useful workflows
  • Interest in building durable internal capability

Probably not the best fit

  • Pure staff augmentation with no product ownership
  • Lowest-price vendor selection
  • Meeting-heavy coordination roles
  • Projects with no clear user or business problem
  • Requests to add AI without a useful workflow

Start a conversation

Let’s discuss the role or the problem.

Start with a short introduction.

Hiring a technical leader, introducing AI or looking for a fractional CTO? Tell me about the company, the opportunity and what you need someone to own.

I will reply with an initial point of view and the most useful next step.

LinkedInConnect professionallyOpens in a new tab GitHubSee public codeOpens in a new tab ToptalView vetted profileOpens in a new tab

Your details stay private. No mailing list and no spam.

Discuss a role or project Clicky