About

A sharp team that ships AI to production

Formentia is a team of AI engineers and researchers who got tired of watching great ideas die in proof-of-concept purgatory. We turn AI from a buzzword into working, deployed systems that create real value.

Founders

Meet the team

Three engineers who have built, scaled, and shipped production software at companies you've heard of. Now we're building AI systems for teams that want the same.

Nate Grover

Co-Founder & Managing Partner

Nate is Co-Founder and Managing Partner at Formentia, where he leads client strategy, engagement delivery, and the firm's overall direction. He spent five years as Director and Principal Architect at Charles Schwab leading cross-enterprise platform architecture, and three years as Senior Principal Solution Architect at Dell EMC driving cloud-native modernization for enterprise customers. Before that he founded Simple Dynamics, a Bay Area consultancy he ran for 15 years delivering platform engineering for clients including WorldRemit, where he led the re-engineering of their global payments platform.

Barrett Tucker

Co-Founder & CTO

Barrett is Co-Founder and CTO at Formentia, where he leads AI engineering, architecture, and technical delivery across client and in-house products. He previously served as Distinguished Engineer and Chief Architect at Charles Schwab, leading enterprise architecture across a platform serving millions of users through 10× transaction growth and the TD Ameritrade integration. Over 20+ years he has architected mission-critical systems across financial services, aviation, retail, and consumer AI for companies including Disney, Nike, and WorldRemit.

Dennis Horte

Co-Founder & Head of Trust & Safety

Dennis is Co-Founder and Head of Trust & Safety at Formentia, where he leads evaluation, red-teaming, and abuse-resistance for the AI systems we ship. He spent a decade at Pinterest, most recently as Engineering Manager for Anti-Abuse, where he led teams across machine learning, signal development, internal tooling, and data science focused on content enforcement and adversarial actors. Prior to Pinterest he was Engineering Manager at Fleksy, leading the engine team behind the keyboard's auto-correction and language models.

Our story

We started Formentia because the AI consulting market is broken. On one side, strategy firms that can't write code. On the other, dev shops that treat AI as just another API call. Neither gets you to production.

We built a different kind of team: engineers and researchers who understand the theory and ship the code. People who can debate model architecture in the morning and have the production infrastructure running by evening.

To prove our depth, we built our own products. MyCloset pushed us deep into multimodal LLM classification, recommendation engines, and computer vision pipelines. Mirror taught us about on-device sensor processing and privacy-first architecture. These projects aren't just portfolio pieces -- they're production systems that real people use every day.

That experience is what we bring to every client engagement: hands-on knowledge of what it takes to get AI from prototype to production.

How we think

Principles that guide every engagement and every line of code.

Engineering depth

We're practitioners, not slide-deck consultants. Every recommendation we make is grounded in real implementation experience shipping AI systems to production.

Pragmatic AI

We match the right tool to the problem. Sometimes that's a frontier LLM, sometimes it's a rules engine. We optimize for outcomes, not hype cycles.

Ship, then iterate

AI projects fail in PowerPoint. We get working systems in front of real users fast, then improve based on data and feedback loops.

Full-stack ownership

From model selection through production infrastructure, we own the entire delivery. No handoff gaps between data science and engineering.

Technical expertise

Deep, hands-on experience across the AI and product stack.

LLM Frameworks

Genkit, LangChain, Semantic Kernel, and direct model APIs. We build structured, evaluated LLM pipelines with prompt management, output parsing, and cost controls.

Agentic Systems

Multi-agent orchestration, tool use, planning loops, and human-in-the-loop workflows. We design agents that are reliable enough for production autonomy.

Recommendation & Personalization

Hybrid recommendation architectures combining embedding retrieval, contextual signals, and LLM reasoning for deeply personalized experiences.

Computer Vision

Image classification, object detection, segmentation, background removal, and multimodal understanding. Cloud and edge deployment experience.

Cloud & Infrastructure

Google Cloud, Firebase, Cloud Run, and serverless architectures optimized for AI workloads. CI/CD pipelines with model evaluation gates.

Product Engineering

Next.js, React, TypeScript, PWA, and cross-platform mobile. We build the full product around the AI -- not just the model.

Want to work together?

We're always interested in ambitious AI projects. Tell us what you're building.

Start a Conversation