20+ years operator experience
Growth, product strategy, marketing operations, automation, and entrepreneurship.
AI infrastructure / Agent memory / Operator systems
Trent Doney
AI infrastructure builder. Memory-first agent systems. 20+ years operator experience.
I build AI infrastructure with an operator's bias: systems need memory, metrics, handoffs, recovery paths, and clear business usefulness. That lens comes from 20+ years across growth, ecommerce, performance marketing, automation, product strategy, and founder/operator work.
Tempe, AZ / Remote
AI infrastructure consulting
Technical co-founder opportunities
$40M+ annual Shopify portfolio
Executive operating context across an eight-store ecommerce portfolio.
5x+ ROAS at $2M+ spend
Performance marketing systems across seven-plus acquisition channels.
47,869 BrainCore facts
Evidence-backed operational memory with 13,465 tracked entities and 265,970 evidence segments.
Operator Profile
AI systems shaped by business judgment.
I build AI systems from the operator's side of the table: infrastructure that has to run, remember, recover, and produce useful work. My background spans growth, ecommerce, performance marketing, automation, product strategy, and entrepreneurship, but the current priority is the local AI operating system running across SynapseGrid Ops.
That system combines an application and automation host, a primary local AI engine, private runbooks, service memory, retrieval, orchestration, agent handoffs, and GPU-backed image and model workflows. It is not a demo stack; it is a working infrastructure layer for building, testing, routing, and improving AI-assisted operations.
Server 1
Primary App Host + Worker Node
- CPU
- AMD Ryzen 9 5900X
- GPU
- RTX 3090 FE 24GB VRAM
- Memory
- 64GB
- Storage
- 3 TB SSD 29 TB archive storage
Runs production apps, automations, worker jobs, media tooling, and always-on services.
Server 2
Primary Local AI Engine
- CPU
- AMD Ryzen Threadripper PRO 9975WX
- GPU
- RTX 6000 Blackwell 96GB VRAM
- Memory
- 128GB
- Storage
- 13 TB SSD 10 TB archive storage
Powers the local LLM stack, image generation, heavy GPU work, and AI production runtime.
Razer Blade Mobile Workstation
Portable AI development, browser automation, client review, Codex control, media ops, and on-the-go orchestration for the local SynapseGrid environment.
- CPU
- Intel Core i9-14900HX 24 cores / 32 threads
- GPU
- RTX 4080 16GB VRAM
- Memory
- 64GB RAM
- Storage
- 1TB SSD
Public Work
The work is visible where the systems live.
github.com/trentdoney
Personal engineering surface for AI infrastructure, memory systems, orchestration tools, and production experiments.
Open GitHubgithub.com/SynapseGrid-Labs
AI infrastructure lab for memory-first agents, orchestration, local AI workflows, and operational systems.
Open SynapseGrid LabsPrivate operating layer
Operational backbone for runbooks, incident memory, service context, retrieval, and repeatable recovery workflows.
Discuss public-safe detailsSelected Impact
Systems work with operational proof attached.
Plus 909 published memories, 1,267 episodes, and 265,970 evidence segments.
Archive-first preservation, extraction, trust classes, and MCP retrieval.
Qwen3-Coder lane with fp8 weights/KV cache in a 96GB VRAM envelope.
Restored execution throughput from 5-7 to 12-18 actions per cycle.
Migrated shared agent infrastructure and wired 60+ agent-environment symlinks.
Improved Recall@10 28.7%, MRR 36.6%, storage 50%, and RSS memory 56%.
Flagship Systems
Production AI infrastructure, not demo theater.
These are not isolated demos. They are connected systems: memory that preserves evidence, orchestration that keeps authority clear, research infrastructure that maps the field, product pipelines that turn signals into workflows, and a private local lab that makes the work repeatable.
BrainCore
Public AI memory infrastructure
Evidence-grounded operational memory for AI agents. BrainCore turns incidents, coding sessions, chats, dashboards, and source changes into queryable memory by archiving artifacts first, extracting facts with provenance, tracking trust and validity, and exposing retrieval through a local MCP-ready layer.
- Facts
- 47,869
- Entities
- 13,465
- Evidence
- 265,970
AgentFanout
Private SynapseGrid Ops orchestration layer
Provider-agnostic routing for bounded multi-agent work. AgentFanout decides when to fan out work, which provider or role should handle it, and when validation is required.
The main session keeps authority over secrets, private tools, git state, destructive actions, and final synthesis. Workers receive bounded packets and return reviewable outputs.
Agent Memory Atlas
Public research corpus
A public map of the agent-memory field: papers, repositories, benchmarks, taxonomies, product docs, protocols, and implementation patterns organized into a structured corpus.
The goal is to help builders compare architectures, understand verification status, and avoid designing memory systems from scattered claims.
ShockFeed
AI product infrastructure
AI product infrastructure for market-signal workflows. ShockFeed turns SEC filing and market signals into ingest pipelines, scoring queues, dashboards, alerts, and repeatable operator workflows.
The value is faster filtering, confidence context, and less noise around time-sensitive filing events.
Local AI Lab
Private AI runtime and operating vault
Local inference, image generation, LoRA training, OCR, captioning, and long-context coding run beside the private OpsVault layer.
OpsVault preserves infrastructure decisions, service context, incidents, remediations, and searchable project memory so agents recover context instead of rediscovering it.
Operating Evidence
The career bridge is the proof.
Operator Track Record
Two decades of marketing and growth work before the AI infrastructure layer.
Before building agent memory and AI infrastructure, I spent 20+ years in growth, ecommerce, marketing operations, product strategy, lifecycle automation, paid media, and founder/operator roles. The through-line is systems that turn messy business motion into measurable execution.
The AI layer is stronger because it sits on that marketing base: attribution, conversion discipline, creative testing, funnel math, customer behavior, and the habit of tying tools to outcomes instead of novelty.
Eight-store Shopify operating context at Interactive Life Forms.
Seven-plus acquisition channels at Scalpa while sustaining 5x+ ROAS.
Particular Paws scaled as a direct-to-consumer brand before a successful exit.
Cash Is King Marketing software and operational growth systems.
Operating Surface
The stack spans agents, memory, model serving, and growth systems.
Experience
A career built around systems that produce outcomes.
BrainCore / Public AI Memory Infrastructure
Designed and shipped a TypeScript + Python memory system that turns operational artifacts into searchable facts, project timelines, patterns, and remediation playbooks.
AI Infrastructure and Product Systems
Built and operated AI production services across model serving, ingestion, monitoring, output quality, incident recovery, Skills, MCP, and agent runtime integration.
Chief Marketing Officer / Interactive Life Forms
Managed eight Shopify stores generating $40M+ annual revenue and optimized multi-team operating execution.
Chief Marketing Officer / Scalpa
Directed $2M+ annual ad spend across 7+ channels while sustaining 5x+ ROAS.
Founder and CEO / Particular Paws
Scaled a DTC brand to $5M+ revenue and exited successfully.
Owner and CEO / Cash Is King Marketing
Built marketing software and operational growth systems tied to $10M+ combined sales.
Trust Layer
Built for systems that need to be trusted after the demo.
Evidence before memory
Operational artifacts are preserved first, then converted into facts, project timelines, patterns, and remediation playbooks.
Role boundaries for agents
Agent work is routed through clear scopes, provider selection, validation, human gates, and reviewable handoffs.
Recovery paths by design
Incidents, service catalogs, runbooks, and retrieval traces become reusable context instead of one-off troubleshooting.
Public / Private Boundary
Public where useful. Private where necessary.
Some systems are public proof surfaces: BrainCore, AgentFanout, and Agent Memory Atlas can be explained, inspected, and improved in public. Others are intentionally private: Local AI Lab and PAI/OpsVault contain operational context, device details, service catalogs, incidents, credentials-adjacent configuration, and client-sensitive workflows.
The public line is simple: publish architecture, principles, verified metrics, sanitized examples, and reusable patterns. Keep private infrastructure, secrets, raw operational history, unreleased strategy, and sensitive incident detail out of the public site.
Work With Me
Best fit: serious AI systems tied to real operations.
Agent memory and retrieval systems
Operational memory, provenance, temporal context, hybrid retrieval, MCP access, and source-bound claims.
Multi-agent routing and handoffs
Provider-agnostic orchestration, worker roles, validation loops, human-in-the-loop gates, and auditable execution.
AI product infrastructure
Scoring pipelines, signal intelligence, dashboards, queues, local generation workflows, and business-operating constraints.
Contact
Useful conversations start with the system, not the pitch.
Best fit: agent memory, AI infrastructure, orchestration, local AI workflows, technical co-founder work, and product systems where reliability and business judgment matter.