Benton RobertsMachine Learning Platform Engineer, MLOps Practitioner, and Cloud Infrastructure Veteran |
benton@bentonroberts.com
|
Profile
Enabling Production-Grade Machine LearningI build and operate the platform layer beneath production machine learning — feature pipelines, reproducible training, low-latency model serving, drift monitoring, and the multi-tenant infrastructure that bridges data-science experimentation to engineered systems. Behind that ML platform work is twenty years of cloud and container engineering in regulated SaaS: the hardcore instincts of an old-school sysadmin, the state-of-the-art practices of an AI-accelerated programmer, and the pragmatism of shipping through IPO, acquisition, and scaling from dozens of VMs to thousands.
Try a live demo of: The Daily Orbit
- solves a difficult business problem for a household name company
- aggregates information from multiple sources in a sophisticated ingest pipeline
- enhances the dataset using battle-tested information retrieval techniques
- provides a chatbot-style interface for gaining competitive insights
- built in under three weeks using rock-solid engineering practices.
Hands-On Skills
Model Training & Feature Pipelines
- Reproducible training pipelines: Airflow DAGs, DVC-versioned data, MLflow tracking, lockfile-pinned dependencies
- Feast feature stores with batch + Kafka-streamed updates, offline ↔ online parity, and active training-serving skew detection
- Containerized training jobs with deterministic seeding and smoke tests that catch silent regressions
Model Serving & Evaluation
- KServe / BentoML / SageMaker serving with shadow deployments and A/B test harnesses
- Multi-axis offline evaluation — NDCG, recall, coverage, novelty — wired to a versioned model registry
- vLLM- and Bedrock-served LLM inference and RAG pipelines (dailyorbit.org)
Inference Monitoring & Drift
- Evidently-driven data, prediction, and concept drift detection
- Prometheus + Grafana dashboards for ML-specific SLOs: feature staleness, hit-rate decay, p99 latency
- Postmortem-driven incident response on ML failure modes (skew, drift, stale features, latency cliffs)
Deeper Expertise
Cloud + Container Engineering
- Deep AWS and infrastructure-as-code expertise
- Kubernetes and container orchestration at production scale, including stateful and GPU-bound workloads
- Cloud-parity fluency on SageMaker and Vertex AI training jobs
Production Reliability
- Scaled a regulated SaaS from dozens of VMs to several thousand, through IPO and acquisition
- Immutable-artifact deployment workflows with full lineage and tested rollback paths
- Deep-dive troubleshooting across the stack: browser → application → kernel → container → GPU
Platform Multi-tenancy
- Onboarded a research-engineering workload (bioacoustic foundation-model fine-tuning) onto an existing ML platform geared for recommendation systems
- Built self-service training, feature, and registry primitives that bridge data-science experimentation to engineered production
- Pre-sales, technical mentoring, and vendor negotiation across multiple senior-staff roles
Recent Experience
2024-2026
Managing Director, Cloud Services, Elevate, Inc. (Denver / Phoenix / D.C.)
- Built AI-enhanced solutions to support eDiscovery, compliance, and other legal services.
- Worked on product development, deep coding, plus infrastructure and identity management.
- Non-technical duties included technical pre-sales / marketing, and leadership / mentoring.
2022–2024
Deputy CTO, Redgrave Strategic Data Solutions (San Francisco / Denver / D.C.)
- Designed and implemented bespoke software solutions to support legal operations.
- Technical duties included a mix of case work, infrastructure development, and product development.
- Non-technical duties included vendor vetting and management, technical pre-sales / marketing, and leadership / mentoring.
2011–2021
Principal Staff Cloud Engineer, Medidata Solutions, Inc. (New York / San Francisco)
- Designed and implemented several different deployment mechanisms based on evolving business requirements. Adopting container-based deployment system saved the company 5-figures per month in AWS costs through increased server density.
- Played a key role in managing the company’s cloud footprint, as it grew from several VMs to several thousand, in a highly regulated industry
- Involved in negotiating and evangelizing company technical standards, from post-IPO to recent company acquisition.
Education
- Dartmouth College, Computer Science
- Columbia University, Mathematics