Benjamin Julia

Product Manager - AI reliability, humans in the loop

From automated evaluation pipelines to projects that failed, this is where I share the products I'm working on and how they're shaping my thinking about AI. Everything here is live : Click, use, judge !

ProjectJanuary 2026

Call Builder — Demystify AI Inside Real Software

A single use case walked step by step through the different layers of AI integration: from a plain prompt up to MCP.

FormationLLMWeb

ProjectJanuary 2026

Deck DSL — Presentation Diagrams, From Plain Text

An extended-Markdown DSL, my own way, to produce the diagrams I need in a presentation — shareable, exportable, and AI-drivable.

DSLAIToolingWeb

ProjectJanuary 2026

JSON Nebula — A Complexity Explorer

When JSON payloads get unreadable or tangle together, JSON Nebula shows them as nodes, links them by equivalent concepts, and tries to give a global view.

Data VizJSONGraphWeb

ProjectJanuary 2026

TableTalk — Talk to Your Database in Plain Language

Import your SQL and chat with your database live, in plain language — and walk away with the code to rebuild it yourself.

FormationLLMAgentSQL+1

ProjectJanuary 2025

Kappa — Usable Signal From Unstructured Data

A platform that tries to pull usable signal out of unstructured data with AI — and to make its reliability measurable through an agreement score, self-calibrating and watching its own drift.

EvalLLMDataAutonomy

ProjectJanuary 2025

madeof.wtf — Made of Chaos

One prompt in, and the rest follows: a character with its 3D model, its spells and its lore — meant to be yours and to follow you from one game to the next.

Game DevAI3D

ProjectJanuary 2025

System Dynamics — Model and Simulate Dynamic Systems

A small, shareable tool to model dynamic systems — stocks, flows, loops — and simulate how they evolve over time, editable by hand or by AI.

SimulationDSLAIWeb

ProjectJanuary 2025

Tinderize — Swiping to Hand Judgment to an AI

An API to spin up swipeable forms in two clicks, results streamed live — on the bet that swiping is one of the fastest gestures to hand an already-formed judgment to a machine.

APILabelingEvalWeb

ProjectJanuary 2023

CRem — Resellers' CRM/ERP (and My AI Lab)

A reseller CRM/ERP I built for my partner, running in her shop for several years — and the ground where I test AI in a real system, not a demo.

CRM/ERPAIEvalUX