What is his strongest engineering work?

Architecting the agent infrastructure behind Fabric, Farpoint's agentic coding IDE: MCP layers for structured tool invocation and multi-step orchestration, tool-execution and memory pipelines, and distributed pipelines that analyze and transform large, multi-file codebases. Earlier, as a software engineer at Cavista in healthcare, he built C#/.NET claims-parsing, optimized SQL, and shipped tested, reliable releases.

What is his full-stack experience?

He ships end to end in TypeScript, from a cross-platform Electron + React desktop IDE (Fabric, at Farpoint) to the web. On the web: React/Next.js front ends, typed APIs, and the data layer with React Query, Zod, and Playwright (this site runs on that stack). His backend history spans C#/.NET at Cavista and Laravel/PHP APIs at Azul, plus a school-management platform for 100,000+ users on Next.js, Node, Docker, and AWS.

What are his credentials?

Best Graduating Student of the School of Computing & Engineering Sciences at Babcock University (First Class, 4.88/5.0, top 1%), an MSc in Computer Science at Ontario Tech University on a Dean's Graduate Scholarship, a peer-reviewed publication at PST 2025 (IEEE Xplore), a $20,000 MITACS research award, and AWS Machine Learning Specialty certification.

What systems does he build under AI agents?

The infrastructure that lets language models plan and act on real code: agent orchestration loops, tool-execution layers, Model Context Protocol (MCP) clients, runtimes, and memory, designed for structured, multi-file work on real repositories rather than one-shot prompting.

What do colleagues say about working with him?

Two named LinkedIn recommendations from his Cavista team. A Senior Engineer noted his skills 'often rivaled those of more senior engineers'; a Product Director noted he 'actively participates in discussions, guiding the team toward optimal decisions' and reviews work thoroughly to fully understand intent.

> Systems & Full-Stack Engineer

IbuildthesystemsunderAIagents(orchestration,toolexecution,MCP,runtimes,memory)andshipfull-stackproductendtoend.

I build the platform layer under AI agents, and the product on top of it.

> Open to full-time systems / platform and full-stack roles. Available for contract on a limited basis.

Get in touchResume GitHub

system.graph

aghoghomena.com agent shell v1. Type `help`.

> whoami

Aghoghomena Akasukpe, systems & full-stack engineer. I build the infrastructure under AI agents and ship full-stack product end to end.

Currently

Agentic AI Systems Engineer, Farpoint (Fabric)

Shipped

Software Engineer, Cavista (healthcare)

Foundation

Best Graduating Student · First Class 4.88/5.0

Research

MSc CS · PST 2025 (IEEE) · $20K MITACS

01How I build// how-i-build

How I build, and what I've shipped

Three things I do well, each grounded in real work: the agentic-coding infrastructure at Farpoint, the healthcare backend at Cavista, and full-stack product across the stack.

Systems & platform

The infrastructure under AI agents.

The runtime that lets agents plan, call tools, and act on real code. At Farpoint I architect the agent infrastructure behind Fabric, an agentic coding IDE: MCP for structured tool invocation and multi-step orchestration, tool-abstraction and execution layers, context-lifecycle and memory pipelines, and distributed pipelines for large-scale codebase analysis and autonomous code improvement.

MCPOrchestrationTool executionMemory pipelinesDistributedTypeScript

Full-stack product

Shipped end to end, desktop to web.

Fabric, Farpoint's agentic coding IDE, is a cross-platform Electron + React + TypeScript desktop app: a Monaco editor, Tree-sitter parsing across a dozen languages, embedded terminals (xterm + node-pty), database awareness, and multi-provider LLM adapters, tested with Vitest and Playwright. On the web I've shipped a school-management platform for 100,000+ users (Next.js, Node, Docker, AWS), this site (Next.js 15, React Query, Zod, Playwright), and apps in Laravel and Django.

ElectronReactTypeScriptMonacoNext.jsNode

Backend & data

Made it fast and correct.

At Cavista I built healthcare-claims parsing in C#/.NET, optimized SQL with EF/LINQ, and handled race conditions for an 80% process improvement, with unit and integration tests and production log analysis. Solid CS fundamentals, kept sharp with competitive coding.

C#/.NETSQL ServerEF/LINQPostgresDockerAWS

How I work

>I dig into the hard problem and ship reliable solutions; colleagues have said my work “often rivaled more senior engineers.”
>I review thoroughly and ask the questions that surface intent before building, not after.
>I document and communicate so the whole team moves faster, and I own the outcome.

02About// about

Short version

Full background, experience & honours >

I'm a systems and full-stack engineer. I build the infrastructure that lets AI agents plan, call tools, and act on real code (agent orchestration, tool execution, MCP clients, runtimes, memory), and I ship full-stack product end to end. Right now I'm the agent-infrastructure engineer behind Fabric, Farpoint's agentic coding IDE: MCP for tool invocation and orchestration, tool-execution and memory pipelines, and distributed pipelines for large-scale codebase analysis and autonomous code improvement.

I ship across the stack. This site runs on Next.js 15, React, TypeScript, Tailwind, React Query, Zod, and Playwright. Before grad school I was a software engineer at Cavista in healthcare (C#/.NET claims parsing, SQL optimization, and tests) and a full-stack engineer building APIs and product in Laravel and Django. The foundation is solid CS: Best Graduating Student of my school at Babcock University (First Class, 4.88/5.0, top 1%) and an MSc in Computer Science at Ontario Tech University on a Dean's scholarship.

Full background & experience Research GitHub

03Selected work// selected-work

Featured work

A couple of representative builds. Each writeup: the context, the constraint, the decisions, and what I would do differently — sanitized where under NDA.

Farpoint Technologies (Fabric)details under NDA

Agent infrastructure for Fabric, an agentic coding IDE

Fabric is an agentic AI coding IDE that plans first and codes second. I architect the infrastructure underneath: how agents invoke tools, hold context, and act on large, multi-file codebases.

outcome

Agents work on real repositories through structured tool calls, persistent context, and distributed analysis, not brittle one-file text edits. Metrics under NDA.

agent-systemsmcpplatform

read the writeup

Cavista Technologies (healthcare)

Healthcare claims parsing and SQL optimization at Cavista

A C#/.NET claims pipeline had to parse diverse Health Care Claim (837) data correctly and fast, with reliable releases across environments.

outcome

Race-condition handling improved a key process by ~80%; SQL optimization and a thorough test suite kept releases reliable across dev, staging, and production.

backenddotnetsql

read the writeup

See all work(4)

04How an agent loop works// agent-demo

How I structure the loop, and how I build it

A steppable plan / tool / result / reflect run over a sample repo (canned, not a live model). It shows how I structure orchestration, route tool calls, and use the reflect step to keep the system observable and recoverable, the same patterns behind the agentic-coding work.

watch an agent work: step 1/8readyfake-repo @ main

planGoal: extract the auth logic out of the request handler into a typed service. Plan: locate the handler, map its call sites, refactor in safe passes, and prove it with tests.
toolgrep -r 'login' src/ → reading the handler and its imports
resultFound 1 handler, 3 call sites, and an existing test file. The handler mixes parsing, auth, and the response.
reflectThe seam is the auth step. Plan holds, but do it in two passes: extract the service first, then move the call sites, so each step stays green.
planExtract AuthService behind a typed interface; keep the handler delegating to it; change nothing observable to callers.
toolwrite src/auth-service.ts · patch handler + 3 call sites · run tests
resultTests: 24 passed, 0 failed. Behaviour identical; the handler is 40% smaller and the auth logic is unit-tested in isolation.
reflectDone. The loop stayed observable: each tool result fed the next decision. What I'd watch: one call site has a subtle default worth a regression test.

now: plan

05Writing// writing

Notes on building systems and shipping product

Full index & RSS >

06FAQ// faq

Questions, answered directly

The things people ask before reaching out.

>What roles is Aghoghomena Akasukpe open to?: Full-time systems / platform engineering and full-stack product engineering roles. He builds the infrastructure under AI agents (orchestration, tool execution, MCP clients, runtimes, memory) and ships full-stack product end to end. He is also available for contract on a limited basis.
>What is his strongest engineering work?: Architecting the agent infrastructure behind Fabric, Farpoint's agentic coding IDE: MCP layers for structured tool invocation and multi-step orchestration, tool-execution and memory pipelines, and distributed pipelines that analyze and transform large, multi-file codebases. Earlier, as a software engineer at Cavista in healthcare, he built C#/.NET claims-parsing, optimized SQL, and shipped tested, reliable releases.
>What is his full-stack experience?: He ships end to end in TypeScript, from a cross-platform Electron + React desktop IDE (Fabric, at Farpoint) to the web. On the web: React/Next.js front ends, typed APIs, and the data layer with React Query, Zod, and Playwright (this site runs on that stack). His backend history spans C#/.NET at Cavista and Laravel/PHP APIs at Azul, plus a school-management platform for 100,000+ users on Next.js, Node, Docker, and AWS.
>What are his credentials?: Best Graduating Student of the School of Computing & Engineering Sciences at Babcock University (First Class, 4.88/5.0, top 1%), an MSc in Computer Science at Ontario Tech University on a Dean's Graduate Scholarship, a peer-reviewed publication at PST 2025 (IEEE Xplore), a $20,000 MITACS research award, and AWS Machine Learning Specialty certification.
>What systems does he build under AI agents?: The infrastructure that lets language models plan and act on real code: agent orchestration loops, tool-execution layers, Model Context Protocol (MCP) clients, runtimes, and memory, designed for structured, multi-file work on real repositories rather than one-shot prompting.
>What do colleagues say about working with him?: Two named LinkedIn recommendations from his Cavista team. A Senior Engineer noted his skills 'often rivaled those of more senior engineers'; a Product Director noted he 'actively participates in discussions, guiding the team toward optimal decisions' and reviews work thoroughly to fully understand intent.

07Contact// contact

Let's talk

Open to full-time systems / platform and full-stack roles, and available for contract. Hiring, or just want to talk shop? Use the form, or reach me directly below.

Hiring / peer path, no form: