Skip to content

Alper Onder, London, UK

Software engineer building AI systems and the infrastructure that makes them reliable enough to trust in production.

Co-founder and technical lead at Krynix1, independent assurance for AI agents. Cloud infrastructure engineer at SITA, behind roughly 45% of global aviation operational data exchange.

  • MSc Machine Learning, Distinction
  • Agent memory, evaluation, reinforcement learning
  • PyTorch, TensorFlow, Kubernetes, Terraform
See featured workcontact@alperonder.dev

Selected work

Krynix1

Co-founder and technical lead

Independent assurance and evidence platform for AI agents. Decisions are held in append-only, hash-chained custody, so any later edit or deletion is detectable.

  • TypeScript
  • Python
  • PostgreSQL
  • OpenTelemetry
  • MCP

RefineX2

MSc dissertation research, scored 87%

Verification-guided LLM reasoning. An iterative controller couples a generator with typed verifiers across arithmetic, algebra, and formal logic.

  • Python
  • SymPy
  • Docker
  • Ollama
Evidence custodyAn agent emits a decision, which is appended to a hash chain of four entries. Each entry commits to the hash of the previous entry, so the log is tamper-evident and any later edit or deletion breaks the chain. A verifier independently verifies the chain and returns the result.agentdecisiona3f17c02e91b4d5aindependent verification of the chainappend-only
Fig. 1Evidence custody. How a decision becomes evidence in Krynix. Each entry commits to the hash of the previous one, making the log tamper-evident: any later edit or deletion breaks the chain. The dashed path is independent verification of what the enforcement layer claimed.
All projects and open source

About

I work at the point where AI systems meet the infrastructure that has to keep them honest. Most of what I build is concerned with the same question: how do you show that a system did what it claims to have done.

The research side is where most of my reading goes: memory and context management for language agents, self-refinement, and how much of an agent's competence comes from architecture rather than scale.

Currently co-founder and technical lead at Krynix, and a cloud infrastructure engineer at SITA.

Full background and career record
Alper Onder, software engineer, London

Contact

Open to conversations about AI systems, agentic infrastructure, and the engineering that keeps them honest.

Sources

  1. 1Krynix, live site. krynix.io
  2. 2RefineX, source. github.com/artificialvirus/RefineX
  3. 3Backstage contribution, pull request. backstage/community-plugins #8181
  4. 4Open source work. github.com/artificialvirus

Claims without a marker are stated without independent evidence. That distinction is deliberate.