ML Platform Architect | DevSecMLOps | Agentic Infrastructure
I architect production ML and AI infrastructure for regulated industries at enterprise scale. I specialize in platform engineering, governance-by-design, and DevSecMLOps patterns that enable velocity without compromising security or observability.
- Multi-tenant ML Platforms: Eliminated siloed infra to enable secure, governed model reuse across 10+ engineering teams.
- Impact: Cut time-to-production from 3-4 weeks to 5 days.
- Governance-by-Design: Embedded PII controls and lineage directly into pipelines, replacing manual compliance reviews.
- Impact: Zero audit risk with no loss in developer velocity.
- Unified Observability: Correlated model drift, performance, and infra health to shift from reactive firefighting to proactive reliability.
- Impact: Reduced MTTR by ~90% and prevented customer-impacting outages.
- DevSecMLOps: Aligned CI/CD to OWASP/NIST/CIS standards, making security automatic rather than a post-deployment gate.
- Impact: Near-zero critical vulnerability escape rates without slowing releases.
Building the layer most AI teams build last: governed agent orchestration, cross-agent memory, and structural prompt injection controls to make autonomous AI safe for production.
- Cloud & Infra: AWS (SageMaker, Glue, Lambda), Kubernetes, Terraform, Docker, MicroK8s, Juju
- Data Platforms: Apache Doris, PostgreSQL, DuckDB, Redis, Kafka, Spark, dbt Core, MetricFlow
- ML & AI Tooling: Charmed Kubeflow, MLRun, Feature Stores, Model Registry, Prometheus, Grafana
- LLM Frameworks: LangChain, LlamaIndex, LangGraph, LangSmith, PyTorch, TensorFlow
- Governance: OWASP ML Top 10, NIST AI RMF, CIS Controls v8, MITRE ATLAS, OPA/Rego
- LinkedIn: linkedin.com/in/felix-isaiah
- Writing: AI System Design (Substack)
- Portfolio: felix-mutinda.github.io
"The best platform architecture solves constraints, not just technical problems."