Journal — Elyadata

Journal10 essays

Notes du terrain.

Apr 2026 · engineering · 5 min readFeatured

Agile in the era of AI

AI is reshaping Agile's core assumptions. Implementation is collapsing in duration, and with that, new priorities emerge. The teams that adapt here aren't the ones just moving faster, but the ones also redesigning how they work to absorb that speed.

Archive· 09 / 09

The Curse of Non-Determinism: Putting agentic AI into production

Your AI agent doesn't have an accuracy problem. It has a variance problem, and it's the central engineering challenge of putting agentic AI into production.

The database: the new substrate of AI memory

The center of architectural gravity is shifting. Owning the context layer now means owning how AI reasons, not just what it knows.

How to calculate the real ROI of an AI investment

Most AI ROI calculations look precise but miss part of the picture. A practical framework to model real value, real costs, and what’s often overlooked.

How We Found You: An agentic prospecting Manifesto

We didn’t find the right prospecting tool, so we built one. What it taught us about making AI systems actually work in production

Data architecture: Bridging the gap between IT and data teams.

Data products don’t fail because of tech, they fail at the interface. Why the IT–Data gap exists, and the role that quietly decides if anything ships.

AI architecture is not a model decision. It's an ownership decision.

Choosing the right model won’t give you an edge. What actually compounds over time: process, context, and architectural control.

Parallel Programming: The right way to run tasks concurrently in C#

Your server has 8 threads, but your batch job uses one. What should take seconds stretches into minutes while CPU capacity sits idle. Parallel programming promises massive speedups, if you can avoid the race conditions, deadlocks, and coordination overhead that come with it.

Data observability: When bad Data derails good decisions

Great decisions start with trusted data, and trusted data is a discipline, a practice built into every stage of your pipeline.

Microservices VS Monolith: Why the debate is misleading?

There's a moment in every architecture decision where the theory meets the reality of your team, your scale, and your deadlines. We hit that moment, and what we learned is that the choice between microservices and a monolith is rarely black and white.

Apr 2025

Published

Continuer la lecture