I design and ship autonomous, self-optimizing software systems—from strategy to production. My work reduces operational drag, cuts manual toil, and increases reliability using agentic architectures, rigorous evaluation loops, and observability-by-design.
- Outcome-first: measurable time-to-autonomy, SLO adherence, and failure recovery benchmarks.
- Agentic-by-default: multi-agent coordination with guardrails, audits, and closed-loop learning.
- Production-grade: zero-downtime rollouts, progressive delivery, and resilient data flows.
| Area | What I Deliver |
|---|---|
| AI Automation | Multi-agent orchestration, policy/guardrail layers, automatic evaluation (quality, safety, drift) |
| Back-End Systems | Event-driven services (Go/Node), low-latency APIs, streaming, queues, idempotent pipelines |
| Observability | Metric/trace-first design, anomaly detection, SLO dashboards, self-heal hooks |
| Platform Ops | Docker, Kubernetes, GitOps, progressive rollout, secure Secrets & config |
| Data Layer | Postgres, Redis, Vector DBs, schema/version control, durability & backup strategies |
- Design for autonomy: every workflow has owner agents + watcher agents + evaluator agents.
- Bias for reliability: golden paths, circuit breakers, graceful degradation.
- Evidence over opinion: decisions flow from telemetry, not vibes.
- Tight feedback loops: small changes, measurable impact, fast rollback.
- ⚙️ Orchestration: event buses, fan-out/fan-in, CRON + on-demand triggers, human-in-the-loop only when needed.
- 🧠 Agents: role-specialized, tool-aware, with persistent memory + retrieval, audit trails for every action.
- 🛡 Safety & Quality: policy checks, sandboxing, offline simulation runs, A/B validation against ground-truth.
- 📈 Observability: SLIs/SLOs defined at inception, auto-ticket on breach, self-heal runbooks attached to alerts.
- Define outcomes & SLOs: success metrics upfront (latency, error budgets, automation coverage).
- Architect & simulate: model agents + workflows, run failure simulations and load tests.
- Deliver & measure: ship with progressive delivery and observable impact.
- Iterate toward autonomy: expand agent responsibilities, shrink human touch points.
- Portfolio: https://khorami.dev
- Linkedin: https://linkedin.com/in/khoramii