Skip to content
00Full-stack software engineer

I build products that are useful, fast, and real.

I build full-stack products with authentication, payments, real-time collaboration, and AI-assisted workflows. The strongest signal is public proof you can inspect: live apps, public repositories, and seven merged PRs across OpenAI, MSW, Astro, and comma.ai.

01
4
shipped products
02
7
merged PRs
03
36
public repos
Felmon Fekadu portrait
LocationAlberta, Canada
EducationB.S. Computer Science
FocusFull-stack, real-time systems, AI/ML tooling
SeekingSoftware engineering roles
Engineering profile

Building systems that solve real problems

I’m Felmon Fekadu, a software developer in Calgary building full-stack products, real-time systems, and AI-assisted workflows. I ship code with clear interfaces, tested behavior, and public proof.

My strongest evidence is public and verifiable: TypeJung at typejung.com with auth, Stripe, email offers, and a real paid upgrade; Collab Editor with live sync and persistence; DocAgent Studio with local-first retrieval and citations; and seven merged pull requests across OpenAI, MSW, Astro, and comma.ai.

I’m looking for engineering roles where product judgment and technical discipline both matter: teams shipping real software with high ownership, clear review standards, and room to grow into harder systems work.

How I work
Clear interfaces
Make the product understandable before making it clever.
Real systems
Build for persistence, auth, payments, sync, and failure paths.
Honest proof
Lead with shipped work and label upstream contributions precisely.
Fast iteration
Ship, inspect, and improve instead of polishing in isolation.
Product ownership
Own outcomes, not just isolated implementation tasks.
Learning in public
Use repositories and PRs as public proof of engineering growth.
Featured work

Case studies

Real products, real constraints, real outcomes. Each project is compressed for fast review and points straight to inspectable proof.

Flagship Product

TypeJung

Live assessment product with auth, Stripe, email offers, and AI-assisted reports.

Built a complete assessment product with a multi-step questionnaire, gated reports, account handling, discount email capture, and AI-assisted insights inside a paid product flow instead of a demo.

Shipped publicly at typejung.com with tiered monetization, public source code, and a validated first paid customer flow.
  • Live product at typejung.com
  • 42-question assessment engine
  • Supabase auth, persistence, email offers, and Stripe billing
  • AI-assisted reports inside paid product flow
ReactTypeScriptExpressSupabaseStripeAI reports
Real-Time Web App

Collab Editor

Real-time collaborative editing with sync, autosave, and persistence.

Built a collaborative document editor with live synchronization, rich-text editing, autosave behavior, and a persistent backend instead of a local-only prototype.

Published as a live app with public source code and clear proof of real-time systems work.
  • Live app plus public repository
  • Socket.io synchronization with autosave
  • PostgreSQL persistence through Prisma
  • GitHub OAuth and dual-service architecture
Next.jsTypeScriptSocket.ioTiptapPostgreSQLPrisma
Local-First AI

DocAgent Studio

Local-first document QA with hybrid retrieval and citation-grounded answers.

Built a local-first RAG system combining SQLite FTS5 full-text search with vector embeddings for hybrid retrieval, citation-grounded answers via Ollama, and GraphRAG for entity navigation.

Shipped as an open-source local-first document QA system with hybrid retrieval, citation grounding, and measurable evaluation metrics.
  • Hybrid retrieval: SQLite FTS5 + vector embeddings
  • Citation-grounded answers via local Ollama models
  • GraphRAG entity navigation
  • Offline evaluation for retrieval recall and citation coverage
PythonFastAPISQLiteFTS5EmbeddingsGraphRAG
Developer Tool

msw-inspector

CLI and GitHub Action that finds gaps in MSW mock coverage.

Built an AST-based scanner using ts-morph that parses both MSW handler registrations and API call sites, normalizes the endpoints, and compares coverage. Shipped as a CLI, npm package, and GitHub Action.

Published to npm as msw-inspector-cli, GitHub Action live on Marketplace. Dogfooded on real repos including typejung.com (found 24 unmocked endpoints).
  • Published to npm as msw-inspector-cli
  • GitHub Action live on Marketplace
  • AST-based scanning with ts-morph
  • Dogfooded on real repos — found 24 unmocked endpoints in typejung.com
TypeScriptAST AnalysisCLIGitHub Actionsnpm
Merged PR

openai/openai-agents-python

Fixed race condition and clarified streaming docs in OpenAI's agent framework.

Identified and fixed the race condition in session writes and authored documentation clarifying cancel-after-turn streaming behavior.

Both PRs accepted upstream in March 2026, fixing data loss for concurrent agent workloads.
  • Merged March 2026
  • Fixed data loss race condition in SQLAlchemy sessions
  • Clarified streaming cancel-after-turn docs
  • Accepted upstream in openai/openai-agents-python
PythonSQLAlchemyAgent FrameworksOSS
Merged PR

mswjs/msw

Fixed open handles and type errors in the most-used API mocking library.

Fixed the open handle leak from infinite delays and resolved the RequestHandler type constraint in setup functions.

Both fixes merged and shipped in MSW v2.12.11.
  • Shipped in MSW v2.12.11
  • Used by 200k+ projects
  • Fixed Node.js open handle leak
  • Fixed RequestHandler type constraint
TypeScriptNode.jsTestingOSS
Merged PR

withastro/astro

Fixed language server completion deferral bug in the Astro framework.

Fixed the completion deferral logic in the language server, with additional PRs open for type system, accessibility, and image tooling improvements.

2 PRs merged, 3 additional PRs open across type system, accessibility audit, and image tooling.
  • 2 merged, 3 open PRs
  • Language server completion fix
  • Type system, a11y, and image tooling contributions
  • Active contributor to Astro framework
TypeScriptLanguage ServerDeveloper ToolsOSS
Merged PR

commaai/opendbc & openpilot

Optimized CI and added fuzz testing across comma.ai's automotive stack.

Cached Hypothesis strategies in car model tests to speed up CI, and authored fuzz testing for openpilot TX against panda on replayed routes.

Merged in opendbc with CI speedup; open PR in openpilot for fuzz testing on replayed routes.
  • Merged in commaai/opendbc
  • Open PR in commaai/openpilot
  • CI optimization via Hypothesis caching
  • Fuzz testing for TX against panda on replayed routes
PythonTestingRobotics/AutomotiveOSS
Technical depth

Engineering decisions, not just polished screens

The UI matters, but the work underneath it matters more. These are the patterns I use to keep products stable and credible.

AI as a product capability

Model features are useful when they fit a clear workflow. I prefer bounded product use cases over AI ornamentation.

Example: TypeJung reports
Assessment flow, account state, and billing exist around the report generator.
AI output serves the product instead of replacing the product.
The proof is a live app, not just a prompt demo.
Use models where they improve the workflow, not where they add theater.

Real-time behavior needs discipline

Collaboration features create state problems quickly. The work is keeping sync, autosave, and persistence coherent under real edits.

Example: Collab Editor
Socket-driven updates for shared editing.
Autosave and persistence built alongside the editor experience.
Real-time UX balanced against backend consistency.
Fast interfaces only matter if the underlying state stays trustworthy.

Public review is part of the proof

Open-source contributions force precision. Seven merged PRs across OpenAI, MSW, Astro, and comma.ai — each surviving maintainer review in codebases used by thousands.

Example: OpenAI, MSW, Astro, comma.ai
Race condition fixes, type system bugs, and test infrastructure improvements.
Public maintainer review and accepted merge outcomes across 4 organizations.
Contributions to projects used by 200k+ downstream dependents.
Reviewed code in public repositories is stronger proof than self-description.
Journey

Experience & milestones

Continuous growth through building, learning, and shipping.

2026Milestone

Open to engineering roles

Seeking software engineering roles where product judgment and technical discipline both matter.

2026Project

7 merged PRs across OpenAI, MSW, Astro, and comma.ai

Upstream contributions fixing race conditions, type system bugs, test infrastructure, and developer tooling in codebases used by hundreds of thousands of developers.

2026Project

Shipped 3 products: TypeJung, Collab Editor, DocAgent Studio

Full-stack products with authentication, payments, email offers, real-time collaboration, and local-first AI-powered document QA.

2025-2026Education

Computer Science studies at University of the People

Focused on software engineering, systems, and full-stack development, with Honors List recognition in Term 2, 2025-2026.

2022-2024Work

Safety Watch / Fire Watch at United Safety

High-risk industrial monitoring, atmospheric readings, compliance logs, and coordination with operations teams in northern Alberta.

2020-2022Work

Mobile Security Patrol Officer

Facility patrols, alarm response, digital incident reporting, and dispatch coordination across mobile security operations.

Open source & experiments

Body of work

Public repositories pulled from my actual GitHub profile, showing current product work, AI tooling, and experiments I build in the open.

View all on GitHub
176Contributions this year
7PRs merged
26Repositories
1Stars earned
Let's work together

Looking for an engineer who can build, learn fast, and own outcomes?

I’m looking for engineering teams that ship real products with high standards — where I can contribute meaningful code from day one.

ff
Index · 2026
© Felmon Fekadu — all rights reserved
Locale
Alberta, Canada
Open to relocation / remote-friendly
Built with
Next.js · Tailwind · Framer Motion
Hosted on Vercel · static edge