# Ömer Faruk Koç

Building production AI/ML platforms, agentic systems, retrieval infrastructure and reliable data systems.

## Machine-readable resources

- https://omerfkoc.dev/resume.md — Aggregated professional profile
- https://omerfkoc.dev/projects.md — Engineering projects and evidence
- https://omerfkoc.dev/experience.md — Professional experience
- https://omerfkoc.dev/skills.md — Technical skills
- https://omerfkoc.dev/learning.md — Current learning and engineering directions
- https://omerfkoc.dev/rss.xml — Published engineering notes feed

## Human-readable website

- https://omerfkoc.dev/ — Portfolio
- https://omerfkoc.dev/projects — Projects
- https://omerfkoc.dev/experience — Professional experience
- https://omerfkoc.dev/learning — Engineering learning map
- https://omerfkoc.dev/writing — Engineering writing
- https://omerfkoc.dev/graph — Engineering graph
- https://omerfkoc.dev/resume — Resume overview

## Published writing

- https://omerfkoc.dev/writing/a-dag-is-a-failure-contract — Orchestration tutorials sell the schedule. The parts that decide whether a pipeline survives are retry semantics, write shape, backfill parameterisation and what a half-finished run leaves behind.
- https://omerfkoc.dev/writing/a-confirmation-is-not-a-boolean — Why "yes" is not always approval, and how interruption, suspension, resume, revalidation, and durable state make consequential agent workflows safe.
- https://omerfkoc.dev/writing/designing-a-policy-driven-model-promotion-control-plane — Model deployment is not complete when a container starts. How progressive traffic, delayed ground truth, explicit policy outcomes, rollback, and desired-versus-observed reconciliation make model promotion a control-plane problem.
- https://omerfkoc.dev/writing/the-write-may-have-succeeded — A timeout after a customer-impacting action does not mean the action failed. How action identity, idempotency, database constraints, and unknown-write handling prevent agents from duplicating business effects.
- https://omerfkoc.dev/writing/memory-is-context-not-authority — Persistent agent memory is useful only if remembered text cannot become permission. A practical boundary for scoped memory, consent, DLP, TTL, and authority isolation.
- https://omerfkoc.dev/writing/testing-ai-agents-without-pretending-they-are-deterministic — Deterministic integration tests and real-LLM tests answer different questions. How to use both to verify control-plane invariants and probe semantic quality.
- https://omerfkoc.dev/writing/decision-authority-execution-observability — For consequential agents, intent, authority, and execution must be observable separately. A practical operator view for workflow state, policy, evidence, replay, and traces.
- https://omerfkoc.dev/writing/rag-can-provide-evidence — Retrieved evidence can support an agent answer without becoming an authorization source. How hybrid retrieval, provenance, grounding checks, and bounded abstention keep the knowledge path separate from execution.
- https://omerfkoc.dev/writing/building-reliable-kafka-event-processing-platform — A deep dive into designing, testing, and optimizing an event-driven commerce platform with Kafka, PostgreSQL, Redis, and transactional outbox patterns.
- https://omerfkoc.dev/writing/63-rescues-0-drops — A cross-lingual reranker moved Recall@5 from 0.9563 to 1.0000. The aggregate is the least interesting number in that sentence, and the latency figure means less than it looks like.
- https://omerfkoc.dev/writing/agent-prompt-injection-guardrails — A defense-in-depth boundary that treats model output, retrieved text, and remembered context as evidence while deterministic software keeps execution authority.
- https://omerfkoc.dev/writing/production-agent-guardrails — A practical execution model for tool-using agents built from typed proposals, deterministic policy, durable confirmation, revalidation, idempotency, and audit.
- https://omerfkoc.dev/writing/rag-citation-integrity — How retrieved source identity, bounded context construction, citation validation, and evaluation turn RAG citations into testable system behavior.
- https://omerfkoc.dev/writing/one-definition-per-feature — When every pipeline rebuilds the same customer feature, the definitions drift apart silently. Making the feature a shared owned artifact fixes that and creates a change-impact problem that column lineage has to answer.
- https://omerfkoc.dev/writing/transactional-outbox-kafka — How the commerce platform commits business state and a Kafka-ready event together, then publishes that event through a recoverable PostgreSQL outbox.
- https://omerfkoc.dev/writing/kafka-at-least-once-idempotency — A concrete Kafka consumer design that combines Redis leases, a PostgreSQL event ledger, transactional writes, bounded retries, and careful offset commits.

## External profiles

- GitHub: https://github.com/negativexq
- LinkedIn: https://linkedin.com/in/omerfkoc