About
Scott Richard
AI Engineering & Infrastructure Consultant
Lancaster, MA
Helping companies design, build, and deploy real-world AI systems — agentic platforms, RAG, and the cloud and infra behind them.
For 17+ years I've built the platforms behind cloud, ML, and AI systems — most recently agentic AI and RAG infrastructure inside FedRAMP-regulated environments at Cisco, GPU-backed ML platforms at Boston Dynamics, and robotics deployments at Berkshire Grey.
// What I specialize in
Where I add the most value
Agentic AI & RAG systems
Production-grade AI agents and retrieval pipelines: LangSmith-style observability, secure tool-call boundaries, evaluation harnesses, and the operational glue that turns a demo into a system you can trust.
ML & GPU platform infrastructure
Kubernetes-based GPU clusters for training and inference, image/data pipelines, autoscaling, and the cost/perf tuning that keeps ML workloads from quietly burning the budget.
Cloud & DevOps platforms
GCP and AWS, Terraform, Kubernetes, ArgoCD, observability with Datadog/Prometheus. End-to-end CI/CD that ships, plus the network and platform plumbing underneath it.
Enterprise compliance for AI
FedRAMP and SOC2-aware architectures: secret management, network segmentation, evidence pipelines, and the auditor-ready paper trail. Hard-won from real audits, not slide decks.
// Track record
Real systems. Real production.
Selected highlights from 17+ years across Cisco, Boston Dynamics, and Berkshire Grey. Not a resume — just where I've done the work that informs how I consult today.
- 01
Cisco
Senior Cloud Engineer (current)Stood up a self-hosted LangSmith observability stack on EKS for AI agent teams, cutting chain-of-thought debugging from hours to minutes. Designed, deployed, and maintained multiple production AI agent pipelines — including an agentic AI compliance pipeline (code → PR analysis → evidence generation) targeting an 80%+ reduction in audit-evidence grooming, and a build-system agent that triages CI failures and recommends fixes to unstick production deploy cycles.
- 02
Boston Dynamics
Senior Staff DevOps/SRE EngineerBuilt and maintained GPU-enabled Kubernetes clusters for ML training workloads, designed network topology for distributed training, and cut compute + egress costs by 20–50% across cloud and hybrid environments.
- 03
Berkshire Grey
Principal DevOps/SRE EngineerOwned end-to-end infrastructure buildout for robotic warehouse deployments — network architecture, BOM, Juniper switching, AWS testbeds via Terraform, EKS environments, and the ELK observability stack tying it together.
- 04
Cisco — early career
Cloud Software & Data Center EngineerBuilt the 18,000-agent contact center test bed from rack-and-stack through production validation. Designed metrics pipelines and Python automation for continuous performance testing across the virtualized data center.
// How I work
What you can expect
Hands on the keyboard
I write the code and design the architecture myself. You're not buying a slide deck or a junior team — you're buying a senior engineer.
Security-aware by default
Years inside FedRAMP and SOC2 environments mean I think about secrets, network segmentation, and audit trails up front, not bolted on later.
Production over prototype
Anyone can build a demo. I build for the day it has to keep running — observability, cost, on-call, the unglamorous parts.
Direct partnership
No account managers, no handoffs. You talk to the person doing the work, every engagement.
// Team & extended network
Senior depth, on demand.
I lead engagements directly. When a project needs depth beyond a single engineer — robotics integrations, distributed data pipelines, ML research, deeper enterprise security — I tap into a vetted network of senior engineers I've worked alongside and mentored over 17 years across Cisco, Boston Dynamics, and Berkshire Grey.
Have an AI or platform problem worth solving?
Discovery calls are free. We'll talk through the problem, what good looks like, and whether I'm the right person to help.