Bartender. Builder. Operator.

The same discipline that makes a great cocktail makes a great system.

I'm Peter Marsala. I read rooms and I read systems. Whether it's a 200-cover Friday night or a 25-system AI platform, I bring the same thing: calm under pressure, sharp priorities, and something that actually works when I walk away.

Explore my work Get in touch

What I bring

Clarity in messy environments
Observable, layer-by-layer systems
AI as specialist, not Swiss army knife
Delivery velocity without chaos

About me

I spent fifteen years behind the bar learning how to read people, manage complexity under pressure, and build systems that work when things get busy. A good bartender knows within thirty seconds what kind of experience a guest wants. That same instinct drives everything I do in technology and consulting.

My path wasn't linear. I studied geology at Western Michigan University, worked in a remote sensing research lab processing satellite imagery, presented at the Geological Society of America, then chose mental health over a PhD. I found my way behind the bar, and it turned out the bar teaches you something academia doesn't: systems exist to serve people, not the other way around.

Today I run Reduct LLC, a consulting practice with two tracks: food & beverage operations and AI strategy. I also built CoachMarsalaCC, a 25-system AI platform I use every day. I don't just talk about AI systems. I live inside one.

Based in Salem, MA

Education BS Geology, Western Michigan University
Lee Honors College

Certifications Google Project Management (2025)
Intuit Bookkeeping (2024)

Currently building

CoachMarsalaCC

Live & in daily use

A 25-system distributed AI coaching and operations platform, built solo from scratch. Not a proof-of-concept — a production system I use every day to manage my work, track patterns, and coordinate AI agents. This is my proof-of-work: I don't just talk about building AI systems, I live inside one.

Agent Bus

28+ MCP tools with event-driven coordination. Multiple AI agents register, discover each other, and hand off work without centralized orchestration.

Coach Decision Engine

Pattern recognition across coaching data that generates proposals for system improvements. Tracks decisions as formal records. 100% approval rate across logged decisions.

Screen State Bus

Real-time screen capture with OCR for observable AI operations. Custom Nemotron OCR service achieving 62ms inference, deployed in Docker.

Living Documentation

Watches the agent bus and automatically enriches sessions with relevant context, known solutions, and recent changes. The system surfaces what's relevant.

AI agents Autonomous ops Observable systems Local-first Docker Discord pipeline

Two tracks, one philosophy

Before automating anything, map the process. Identify where friction lives. Design the improved version. Automation comes after the design is proven manually.

Food & Beverage Consulting

Fifteen years behind the bar taught me how to build repeatable systems under real pressure. I help restaurants and bars elevate guest experience, unlock margin, and create operations that run without the owner in the room.

Multi-site program rollouts across 4 simultaneous client engagements
3-level training programs that cut staff ramp time 25-30%
SOPs, playbooks, and documentation systems built for real teams
Menu engineering, ops design, and guest experience strategy

Ops design Training programs Menu engineering Guest experience

Start a conversation →

AI Strategy & Operations

Through Reduct LLC, I help teams deploy AI that keeps humans in control. Not "we should use AI somewhere" hand-waving — concrete plans: what problem, what tool, what changes to process, what does success look like.

AI workflow design with observable, auditable infrastructure
Process mapping before automation — fix it first, then speed it up
Agent architecture and coordination patterns
Sustainable delivery velocity — ship fast without building debt

Program delivery AI workflow design Agent architecture Process automation

Let's talk →

Where I've been

2022 – Present

Bar & Cocktail Consulting Consultant

Salem, MA

End-to-end program cycles for four client businesses: discovery, requirements capture, delivery planning, rollout, and iterative improvement. Translated client needs into documented plans, training curricula, and standardized operating procedures.

2024 – 2025

Danger Beams Co-Founder

Salem, MA

Laser-cut goods business built from zero. Designed tracking systems, vendor workflows, and financial infrastructure. Standardized and automated recurring tasks to create repeatable, consistent execution.

2021 – 2025

Deacon Giles Distillery Lead Bartender & Bar Manager

Salem, MA

Designed a quarterly 3-level training program for 12-16 participants per cohort. Built documentation systems (SOPs, manuals, playbooks) that survived team changes. Applied operational feedback to reduce staff ramp time 25-30% and contributed to the first fully profitable year under management.

2019 – 2021

Encore Boston Harbor Bartender

Everett, MA

Supported launch operations for two new restaurants in a Forbes Five-Star environment. Day-one readiness, cross-team coordination, and consistent execution within strict service standards.

2016 – 2018

Far From the Tree Cider Brand Rep & Tap Room

Salem, MA

Represented the brand at large-scale events, coordinated logistics and customer engagement. Authored monthly newsletters supporting marketing and customer communications.

2016

CenterWatch Web Content Associate

Boston, MA

Content management and web publishing for a clinical trials information services company.

2014 – 2016

Thermal Form & Function Market Resource Manager

Beverly, MA

Market research and resource management for a thermal management solutions company.

2009 – 2012

Western Michigan University Research & Teaching Assistant

Kalamazoo, MI

Processed satellite imagery and data in a remote sensing and hydrology lab. Wrote scripts to automate raster processing, co-authored publications, and presented at the Geological Society of America on NASA GRACE satellite data.

How I think about work

AI should be a specialist

General-purpose AI is impressive but not very useful. The real value comes from building the right context, tooling, and constraints so it becomes a specialist at a specific job. A bartender with a well-organized station makes better drinks than a genius with a messy one.

Everything observable

No black boxes. If a system is running, you should be able to see what it's doing, why it made a decision, and what happened last time. You can't improve what you can't see.

One layer at a time

Build the simplest thing that works. Make it observable. Then layer on automation only when the manual version proves the concept. This prevents the common failure mode of automating a broken process and making it break faster.

The system serves the person

A perfectly optimized cocktail program means nothing if guests don't feel welcome. A perfectly architected platform means nothing if it doesn't solve a real problem for a real person. The question is never "what's the most elegant architecture?" — it's "what actually helps?"

Get in touch

Best way to reach me: cheers@petermarsala.com

Want to meet AI Peter? He's behind the bar.

Chat with AI Peter →