# Ömer Faruk Koç — Learning & Engineering Directions

## 01 — Agent Systems

Orchestration, deterministic policy and durable confirmation are implemented and evaluated; memory governs what persists across steps and future interactions.

**Current foundation:** Stateful LangGraph orchestration, deterministic policy and persistent memory  
**Next direction:** Measured memory relevance, retention and compaction at scale

### Agentic Memory (learning)

Customer-scoped persistent memory with consent, TTL and deletion already runs in the agentic platform, where remembered text is contextual evidence and cannot authorize work. The open question is how memory behaves as it grows: which entries stay worth retrieving, how retention and compaction should be decided, and how relevance is measured rather than assumed.

**Exploring:** Episodic memory, Semantic memory, Memory retrieval, Relevance measurement, Memory compaction, Retention policies, Memory lifecycle, Conflict handling

**Evidence target:** Extend the existing persistent-memory implementation with a repeatable benchmark for retrieval relevance, retention behavior, compaction and conflict cases as memory volume grows.

**Related projects:** [Agentic Customer Service Platform](https://omerfkoc.dev/projects/agentic-customer-service-platform)

## 02 — Retrieval & Evaluation

Retrieval finds evidence; context engineering selects and budgets it; GraphRAG adds relationships; evaluation measures whether each change helps.

**Current foundation:** Hybrid retrieval, reranking and citation integrity  
**Next direction:** Context construction, GraphRAG and repeatable evaluation

### Context Engineering for RAG (learning)

Extending existing hybrid retrieval, reranking, citation-integrity and source-grounded context work by exploring how evidence should be selected, ordered and compressed before generation instead of passed through as a naive top-k chunk dump.

**Exploring:** Context construction, Chunk selection, Context ordering, Context budgeting, Redundancy reduction, Coverage-aware selection, Diversity-aware selection, MMR, Metadata-aware context, Query-aware context assembly, Source-grounded context, Context compression, Context-window management

**Evidence target:** Build a repeatable benchmark comparing naive top-k retrieval with diversity-aware and coverage-aware context selection across answer quality, context size, source coverage and latency.

**Related projects:** [Knowledge Base RAG](https://omerfkoc.dev/projects/knowledge-base-rag)

### GraphRAG (learning)

Exploring when explicit entity and relationship structure improves retrieval compared with vector-only RAG, especially for multi-hop and relationship-heavy questions.

**Exploring:** Knowledge graphs, Entity extraction, Relationship modeling, Graph traversal, Multi-hop retrieval, Vector retrieval, Metadata filtering, Graph + vector retrieval, Neo4j

**Evidence target:** Build a Neo4j-backed GraphRAG experiment and compare it against conventional hybrid RAG on the same evaluation set.

**Related projects:** [Knowledge Base RAG](https://omerfkoc.dev/projects/knowledge-base-rag)

### LLM / RAG Evaluation (learning)

Extending existing DeepEval, bilingual reranker and citation-validation work into repeatable benchmark suites that measure retrieval quality, reranking behavior, citation integrity, context efficiency, latency and failure modes.

**Exploring:** Retrieval evaluation, Reranker evaluation, Recall@K, MRR, nDCG, Citation integrity, Grounding / faithfulness, Context efficiency, Latency, Failure-mode testing, Reproducible benchmark datasets, Regression testing

**Evidence target:** Create a versioned evaluation dataset and automated benchmark comparing retrieval and context-construction configurations across quality, grounding, context size and latency metrics.

**Related projects:** [Knowledge Base RAG](https://omerfkoc.dev/projects/knowledge-base-rag)

## 03 — Platform Infrastructure

Infrastructure provisioning, workload orchestration and observability extend containerized AI systems along distinct operational boundaries.

**Current foundation:** Containerized services and observable release workflows  
**Next direction:** Remote state, cloud infrastructure and workload orchestration

### Terraform (learning)

Building a production-style local infrastructure lab with modular Terraform, the Docker provider, state migration through moved blocks, native Terraform tests, CI validation and security scanning.

**Exploring:** Infrastructure as Code, Terraform Modules, Docker provider, State migration, Moved blocks, Native Terraform tests, CI validation, Security scanning

**Evidence target:** Explore remote state, environment/state isolation, cloud infrastructure, CI plan workflows and workload orchestration without presenting them as demonstrated capability yet.

**Related projects:** [Terraform Docker Infrastructure Lab](https://omerfkoc.dev/projects/terraform-docker-infrastructure-lab)

### AI Platform on Kubernetes (planned)

Extending containerized AI platform work toward production orchestration, with emphasis on model serving, rollout control, resource isolation, scaling and observability.

**Exploring:** Model serving, Application workloads, Autoscaling, Rollout strategies, Health checks, Resource requests and limits, GPU scheduling, Observability, Service networking, Configuration, Secrets, Workload isolation

**Evidence target:** Deploy an existing ModelOps or RAG workload on Kubernetes with health checks, resource controls, rollout strategy and observable service behavior.

**Related projects:** [ModelOps Control Plane](https://omerfkoc.dev/projects/modelops-control-plane)

### AI Platform Observability (learning)

Extending existing OpenTelemetry, Jaeger, Prometheus, Grafana and rollout-verification work toward AI-platform-specific signals across model serving, retrieval, agent execution and release automation.

**Exploring:** Distributed tracing, Model-serving metrics, Retrieval latency, Token usage, Agent traces, Tool-call traces, Rollout metrics, Failure classification, SLOs / SLIs, Alerting, Evaluation telemetry

**Evidence target:** Define and validate an observable AI service workflow with end-to-end traces, platform metrics, failure classification and a small set of explicit SLIs.

**Related projects:** [ModelOps Control Plane](https://omerfkoc.dev/projects/modelops-control-plane), [Knowledge Base RAG](https://omerfkoc.dev/projects/knowledge-base-rag), [Agentic Customer Service Platform](https://omerfkoc.dev/projects/agentic-customer-service-platform)

## 04 — Software Systems Engineering

Deepening the systems-level engineering foundations behind reliable production AI and distributed applications through failure analysis, measurable performance and explicit service-boundary trade-offs.

**Current foundation:** At-least-once delivery, idempotency, load-tested throughput and transactional service boundaries already implemented and measured  
**Next direction:** Explicit failure-scenario testing, profiling-driven performance work and service-level trade-off benchmarking

### Distributed Systems & Reliability (Deepening)

At-least-once delivery, idempotent consumers, bounded retries and a transactional outbox are already implemented and benchmarked in the commerce platform. The open direction is deeper: reasoning explicitly about partial failure — what a service should do when a workflow fails halfway, when eventual consistency is an acceptable trade-off rather than a shortcut, and where a retry helps versus where it turns a transient failure into a duplicate or a cascading one.

**Exploring:** Queues and asynchronous processing, Retries and retry boundaries, Idempotency, Consistency models, Backpressure, Failure recovery

**Evidence target:** Extend the commerce platform's failure-injection surface with explicit failure-scenario tests — partial workflow failure, duplicate delivery, retry storms and consistency-window violations — each with a reproducible before/after outcome, not just a passing happy-path suite.

**Related projects:** [Real-Time Commerce Platform](https://omerfkoc.dev/projects/real-time-commerce-platform)

### Concurrency & Performance Engineering (Building)

Locust-driven load testing and query-plan-driven latency optimization already produced measured, reproducible numbers on the commerce platform and the ModelOps benchmark suite. The open direction is treating concurrency itself as an engineering variable: profiling to find where time is actually spent, distinguishing CPU-bound from I/O-bound from contention-bound workloads, and using Go's goroutine and channel model as a second concurrency substrate to compare against Python's async/concurrency model.

**Exploring:** Concurrency and parallelism, Go goroutines and channels, Profiling, Benchmarking, Load testing, Bottleneck analysis

**Evidence target:** Produce reproducible benchmarks and profiling reports — before/after comparisons under load, with the saturated resource identified — rather than reporting a single throughput number without its bottleneck.

**Related projects:** [Real-Time Commerce Platform](https://omerfkoc.dev/projects/real-time-commerce-platform), [ModelOps Control Plane](https://omerfkoc.dev/projects/modelops-control-plane)

### Networking & Service Engineering (Exploring)

FastAPI service boundaries, PostgreSQL transactional writes and idempotency-key design are already implemented across the agent, ModelOps and commerce platforms. The open direction moves underneath the endpoint: HTTP connection and timeout behavior, gRPC as an alternative to REST for internal service communication, transaction isolation levels and connection pooling, and the operating-system behavior a production service actually runs on top of.

**Exploring:** HTTP internals, gRPC, Networking fundamentals, Database transactions, Linux

**Evidence target:** Implement, benchmark and document concrete service-level trade-offs — REST vs. gRPC for an internal call path, transaction isolation levels under concurrent writes, connection-pool behavior under load — rather than treating any of them as a fixed default.

**Related projects:** [Agentic Customer Service Platform](https://omerfkoc.dev/projects/agentic-customer-service-platform), [ModelOps Control Plane](https://omerfkoc.dev/projects/modelops-control-plane), [Real-Time Commerce Platform](https://omerfkoc.dev/projects/real-time-commerce-platform)