Context Engineering
Designing information architecture, tool definitions, retrieval patterns, and execution boundaries so agents can operate reliably.
AI Platform Engineer · Context Builder · Product Engineer
I build AI systems that understand, act, and deliver real value.
I design context pipelines, MCP servers, integrations, internal platforms, and customer-facing products that connect models to the information and tools they need.
Looking for Salesforce work? Visit SFdork.io
ouch that smarts
context.window87%
tool.surfaceconstrained
schema.checkpassed
permission.modeleast privilege
10+Years building business systems
AIMCP and agent infrastructure
APIEnterprise integrations
UXEngineering, product, and strategy
āIndependent products released
What I do
Systems that make AI useful outside the demo.
My work sits where models meet messy business workflows: schemas, authentication, permissions, observability, UX, and the product judgment needed to ship.
MCP and Agent Infrastructure
Building secure agent-facing tools for systems such as Salesforce, HubSpot, SFTP, and PubMed with useful schemas and predictable errors.
Enterprise Integrations
Connecting fragmented platforms through APIs, authentication, event-driven workflows, and dependable data movement.
Full-Stack Product Engineering
Taking products from idea through architecture, implementation, UX, release, telemetry, and continued iteration.
Developer and Internal Platforms
Creating systems that reduce friction for engineers, operators, GTM teams, and end users.
Human-Centered AI
Designing workflows where model behavior, permissions, feedback, and failure states stay legible to the people using them.
Featured work
Case studies across agents, systems, and shipped products.
Selected work across agent infrastructure, enterprise systems, and independently shipped products.
Enterprise MCP Platform at Infinitus
Reusable infrastructure for agent access to isolated systems.
Problem
AI agents needed reliable access to business and healthcare-related systems, but each source had different authentication, schemas, permissions, failure modes, and data semantics.
Approach
Built MCP servers and integration layers for Salesforce, HubSpot, SFTP, and PubMed. Designed agent-facing tools around clear schemas, constrained actions, predictable errors, and useful context rather than simply exposing raw APIs.
Outcome
Created reusable infrastructure that allowed agents to interact with previously isolated systems through consistent, governed interfaces.
Lessons
- Tool descriptions and schemas strongly influence agent behavior.
- A larger tool surface does not necessarily make an agent more capable.
- Predictable failures are more useful than opaque success.
Salesforce MCP and Secure Agent Access
Agent tooling, context design, and security boundaries.
Problem
Enterprise teams needed AI agent capabilities in Salesforce without weakening the security, compliance, and data-access boundaries required by regulated environments.
Approach
Designed controlled execution contexts, scoped API access, and agent-facing actions that exposed useful capabilities while preserving governance and auditability.
Outcome
Established reusable patterns for secure AI integration in Salesforce and reduced investigation friction without expanding unnecessary permissions.
Lessons
- Permissions and observability matter as much as model capability.
- Context selection beats dumping every available field.
- Agent tools need product thinking, not just API plumbing.
Agent-Ready GTM Systems
Turning fragmented revenue systems into usable context.
Problem
Customer, marketing, sales, support, and operational data often live across disconnected systems, making it hard for people or agents to understand what matters.
Approach
Designed integration patterns, data models, routing logic, enrichment flows, and automation surfaces that turn distributed GTM activity into reliable operational context.
Outcome
Helped teams act from a clearer shared picture while creating the foundations that make future agent workflows more dependable.
Lessons
- Great AI workflows usually start with coherent source systems.
- Data semantics are product decisions.
- Useful context needs timing, permissions, and ownership.
From Idea to Shipped Product
Independent software across mobile apps, games, and experiments.
Problem
Independent products require more than implementation: discovery, UX, accessibility, distribution, pricing, support, feedback, and iteration all matter.
Approach
Designed and shipped products end to end, including Couch X Crawler, JunkCam, That Counts Today, Full Signal, LexInvaders, and other prototypes.
Outcome
Couch X Crawler reached #2 in the Apple Entertainment chart, and the broader archive demonstrates a steady habit of building, finishing, and learning from real users.
Lessons
- Shipping teaches tradeoffs that prototypes avoid.
- Support and onboarding are part of engineering.
- Small products sharpen product sense quickly.
Project history
Professional work systems and public apps.
A broader archive of professional systems and public products, from enterprise platforms to mobile apps and games.
Showing all 11 projects
Work systems
Professional projects built inside company environments.
Internal platforms, MCP servers, Salesforce systems, and integration work created as part of employed roles.
MCP Infrastructure
Agent-facing tools, schemas, auth boundaries, and reusable integration layers across enterprise systems.
Salesforce MCP
Secure access patterns for agents operating around CRM data, workflows, OAuth, and controlled actions.
HubSpot, SFTP, and PubMed MCPs
Purpose-built tool surfaces for systems with different data semantics, failures, and permissions.
GTM Platform and Integrations
Routing, enrichment, automation, and data models that make revenue systems usable by teams and agents.
Salesforce Layout Manager
Self-service page personalization through Salesforce preferences and a drag-and-drop Chrome extension.
Realtime Workforce Scheduling
Responsive Experience Cloud scheduling with real-time capacity checks and license-aware access.
Independent products
Public apps and games I designed and shipped myself.
Each card links out to the named product site and uses a preview from that site.
Couch X Crawler
Passive dungeon crawler mixed with virtual companion systems, progression design, and mobile release iteration.
JunkCam
A camera product exploring constraints, texture, playful editing, and fast creative loops.
That Counts Today
A focused utility product shaped around clear feedback, habit loops, and accessible everyday use.
LexInvaders
A retro word-shooter experimenting with randomness, difficulty curves, and repeat engagement loops.
Full Signal
A product effort around interface clarity, platform distribution, iteration, and user trust.
Experience
Support to systems ownership to AI infrastructure.
The narrative is a progression: understand the business, own the platform, then build the context and tools agents need to do useful work.
Infinitus
Applied AI Engineer / Forward Deployed Engineer
MCP development, agent workflows, enterprise integrations, full-stack product work, and GTM systems at Infinitus.
Block / Cash App
Platform Engineer, Salesforce Focus
Salesforce-centered platform engineering, secure enterprise systems, automation, internal tooling, incident workflows, and scaled platform development.
PandaDoc
Senior Salesforce Administrator
Revenue systems, process design, integrations, automation, and stakeholder-facing platform ownership.
Earlier roles
Community banking and CDK Global
Customer-facing support, process fluency, systems thinking, and the foundation for business-platform engineering.
How I work
A practical loop for real systems.
Understand the system
Map users, data, tools, constraints, permissions, and desired outcomes before choosing a technical approach.
Shape the context
Decide what information is relevant, how it should be structured, and what an agent or user should be allowed to do with it.
Build for observability
Design predictable behavior, useful errors, traceable actions, and safe failure paths.
Ship and learn
Put systems in front of real users, measure where they fail, and improve through actual usage.
About
Systems, products, and people.
I am an engineer who enjoys the space between systems, products, and people. My career began in platform support and business systems, grew into Salesforce engineering and enterprise architecture, and now centers on MCP infrastructure, AI agents, integrations, and full-stack product development.
I am most interested in problems where the challenge is not simply generating an answer, but giving a system the right context, tools, permissions, and feedback to accomplish something useful.
Outside of work, I independently design and ship mobile apps and games. That keeps me close to users and forces me to think beyond implementation: product decisions, accessibility, onboarding, distribution, support, and iteration are all part of the job.
Contact
Let's build something thoughtful.
Whether you are hiring, collaborating, or just want to talk AI systems, Salesforce architecture, design, or independent products, I am open to a good conversation.
