Skip to content
View felmonon's full-sized avatar
5 followers · 9 following

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx2Achievement: QuickdrawAchievement: YOLO

Achievements

Achievement: Pair Extraordinairex2Achievement: Pull Sharkx2Achievement: QuickdrawAchievement: YOLO

Highlights

Block or report felmonon

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
FELMONON/README.md
Felmon Fekadu builds developer tools and AI reliability systems that leave evidence; an engineering trace passes through observation and verification while a policy gate blocks a regression.

Felmon Fekadu

I build software that leaves evidence.

Developer tools and AI reliability systems that expose hidden behavior, enforce deterministic gates, and make regressions impossible to ignore.

Portfolio · Technical writing · Email · LinkedIn

01 / CLAIM

Most software says “it worked.”

I care about the harder questions: What happened? What changed? What can we prove?

I build the layer that answers those questions.

02 / INSTRUMENTS

Failure mode Instrument Evidence emitted
Mock drift msw-inspector Coverage report + CI gate
Agent trajectory regression agent-reliability-harness Policy findings + baseline diff
Retrieval without evidence docagent-studio Citations + offline evaluation
Opaque coding-agent quality agent-fight-club Scored, replayable bouts
Unproven product assumptions TypeJung · source A live, paid user journey

03 / ONE RECORDED TRACE

A six-step trace imports ambiguous agent behavior, verifies schema, budget, safety, and grounding, compares candidate to baseline, blocks a regression, and emits a reproducible report.

A demo shows that something worked once. A trace helps explain why it will keep working.

04 / OPERATING CONSTRAINTS

01  Evidence over adjectives.
02  Deterministic checks in CI. Model judgment offline.
03  Narrow interfaces. Reproducible failures. Useful artifacts.

05 / PROOF LEDGER

Record Public evidence
PACKAGE / NPM msw-inspector-cli v0.3.2 + GitHub Marketplace Action
PACKAGE / PYPI agent-reliability-harness v0.2.2
UPSTREAM 14 merged pull requests across 7 repositories / 5 organizations
COMMUNITY 5 outside human contributors in msw-inspector
PRODUCT TypeJung, a live full-stack paid product
WRITING Deterministic checks beat model-judged evals in CI

Dated public proof snapshot, verified 2026-08-01. Every count also links to its live public record.

06 / HANDOFF

Have a failure mode you cannot see yet? Send me the trace.

felmon.tech · felmonon@gmail.com · all repositories

CALGARY / CANADA · TYPESCRIPT / PYTHON · PROFILE DATA · NOT A RÉSUMÉ. A TRACE.

Pinned Loading

  1. msw-inspector msw-inspector Public

    Static analysis for MSW: finds drift between real API calls and mocked handlers, locally and in CI.

    TypeScript 5

  2. agent-reliability-harness agent-reliability-harness Public

    Deterministic policy-as-code and regression gates for AI agent tool-use traces.

    Python

  3. agent-fight-club agent-fight-club Public

    Replay-first arena for benchmarking coding agents on correctness, cost, resilience, and diff quality.

    TypeScript

  4. trace2test trace2test Public

    Turns failed AI agent traces into reproducible regression tests and verified fixes.

    TypeScript

  5. docagent-studio docagent-studio Public

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

    Python

  6. jungian-typology-assessment jungian-typology-assessment Public

    A shipped full-stack assessment product with auth, Stripe billing, saved results, and AI-assisted reports.

    TypeScript 1