Firm · Information AI · since 2003

We build the AI you own.

Group e-media is an Information AI consulting and R&D firm out of Montréal. We connect data-driven companies to AI — from data foundations to self-optimizing agents — and we transfer the work into infrastructure you own.

What we are

An information AI firm — not a media AI firm. We build the substrate, runtime, eval system, and feedback loop that make AI useful for an actual business operation. Source graphs, agent runtimes, model gateways, MCP tool registries, eval-driven delivery, closed-loop knowledge.

Platform-agnostic, vendor-agnostic, vertical-agnostic. From productized AI surfaces to bespoke infrastructure, we design and hand-off the system — your code, your data, your gateway, your traces — so you keep the leverage when the next model ships.

How we work

Outcome-first

Every engagement starts from a real operational outcome — fewer escalations, faster resolution, a closeable churn signal — and works backward to the substrate that makes the outcome durable.

Owned by you

The code, the gateway, the indexes, the eval set, the traces — all yours at the end. No vendor lock-in, no opaque SaaS, no managed-service tax forever.

Eval-driven

Nothing ships without an eval that catches regressions. Promotion gates, regression sets, gold cases — borrowed from software engineering, applied to AI.

Closed-loop

Failures route into the system as knowledge updates, workflow patches, or new eval cases — so the next encounter is better, not the next pilot.

Bilingual by default

Every deliverable ships EN + FR. Voice, surface, governance — all bilingual without retrofit work later.

Knowledge transfer

We pair with your engineers and explain the system as we build it. By production day, the discipline lives in your team, not in our heads.

Our history · 2003 → 2025 · Twenty-two years

From frameworks to cloud to data to AI.

Each era demanded a different substrate: frameworks, cloud, data, then agents. Group e-media stayed close to the metal and kept shipping systems that had to work after launch.

Era 01 / 06
2003

The frameworks era

Founded as a small studio shipping web work for Quebec SMEs and agencies. Open-source stacks, hand-rolled CMS, deeply opinionated about craftsmanship.

Stack
PHPMVCjQueryBootstrapLinuxMySQL
Highlights
  • Custom CMS
  • Editorial sites
  • Agency delivery
  • Open-source craft
Era 02 / 06
2010

Mobile + commerce

Native apps when iOS/Android still meant building twice. Custom commerce platforms when off-the-shelf wasn't enough.

Stack
iOSAndroidCustom CMSPayment gatewaysResponsive web
Highlights
  • Native mobile
  • Commerce platforms
  • Payment flows
  • Responsive systems
Era 03 / 06
2015

Cloud + modern web

AWS, containers, single-page apps. Production at scale, with infrastructure that didn't fall over.

Stack
AWSDockerReactNodePostgresCI/CD
Highlights
  • Cloud operations
  • Containerized apps
  • Modern frontends
  • Production CI/CD
Era 04 / 06
2018

Data became the product

Warehouses, attribution, telemetry. The shift from "build the website" to "build the system that learns from itself."

Stack
BigQuerydbtAirflowStreaming pipelinesCustom attribution
Highlights
  • Attribution systems
  • Telemetry pipelines
  • Warehouse modeling
  • Learning loops
Era 05 / 06
2023

Agents and the LLM era

Frontier models made agent systems practical. The work shifted to orchestration, tools, memory, tracing, and evals.

Stack
AnthropicOpenAIMCPVisual workflowsMulti-agent routingConversation analytics
Highlights
  • Multi-agent runtimes shipped
  • Workflow execution traces
  • Tool and retrieval observability
Era 06 / 06
2025
Now

Information AI · Now

The thesis: own the substrate, launch the agents, evaluate every change, and close the loop from production back into the system.

Stack
Anthropic / OpenAI / Google / BedrockOpen-weights · local · private inferenceSelf-optimizing agentsConversation intelligenceSynthetic personality (R&D)
Highlights
  • Information AI platform launch
  • Self-optimizing agent research
  • Bilingual EN/FR consulting + lab

Where we work

Montréal, Québec. We work bilingually (EN · FR) and we ship to wherever the data and the team live — Canada, the US, EMEA. On site when the engagement needs it, remote otherwise.

Want to talk?

Tell us about your data, your team, and the outcome you're after. We'll come back with an honest scoping conversation — no deck, no demo.

Founder

Alain Carpentier
Founder, Group e-media · since 2003

Two decades of building information systems and product infrastructure for data-driven companies. The firm started as a media-engineering shop in 2003 and grew into the AI infrastructure work it does now — same operating discipline, new substrate.

« Cracking the code of technology — since 2003. »