AI in production. On your hardware. Under your control.
QATALYSTIC builds the hybrid Kubernetes platform, the data foundation and the applied intelligence that runs on top of it — then hands it over to your team, documented, tested and operable without us.
You have the mandate. The platform underneath it is the problem.
Your estate is hybrid because it had to be — regulated data on your own hardware, elastic workloads in the cloud, and now inference wherever it is fastest or cheapest that week.
Every vendor you have spoken to owns one slice of that and hands the rest to someone else. The result is a platform team, a data team and an AI initiative that each work, and a company where nothing reaches production.
We take the whole line, and we hand it back operable.
Five practices, one accountable line
Platform, data, models and architecture are one problem wearing four job titles. Each practice below is a capability we hold in-house, and most engagements draw on more than one.
Cloud & platform engineering
A platform your engineers can ship onto on the day they join.
Data engineering
One lakehouse your analysts trust, fed by pipelines that tell you when they are wrong.
Data science & machine learning
Models that survive contact with production, and the evidence that they are worth running.
Applied AI
AI that runs in your estate, on your data, with its cost and its behaviour under control.
Architecture & software engineering
Decisions written down, systems that can be changed, and a team that can keep changing them.
We arrive with an architecture, not a blank page
This is the reference stack we build on, and the one our own company runs on. Each layer is delivered declaratively, documented, and handed over with the runbooks for operating it.
Applied intelligence
The models that do the work, with routing, evaluation and guardrails around them.
- Private LLM serving
- Retrieval
- Voice & video
- Agents
- Evaluation harness
Data foundation
The lakehouse and the streaming spine, with semantics that are proven rather than assumed.
- Lakehouse
- Streaming backbone
- Contracts
- Metric layer
- Equivalence lane
Platform
The Kubernetes estate and everything a team needs before it can ship anything at all.
- Clusters
- GitOps delivery
- Observability
- Policy & secrets
- GPU scheduling
Fabric
On-premises hardware, cloud regions and serverless inference, addressed as one estate.
- On-prem bare metal
- GKE · EKS
- Serverless inference
- Zero-trust network
Our own synthetic colleagues, in production
These are not a demo. They answer our phone, transcribe our calls, draft our client reports and review our pull requests — scheduled across on-premises GPUs, cloud clusters and serverless inference, wherever each one belongs.
| Colleague | What it does | Runs on | Requests · 24h |
|---|---|---|---|
Aria · Voice Conciergellama-3.3-70b · Groq LPU | Answers the company line, qualifies inbound work, books discovery calls and hands a written brief to a human partner. | Groq | 1,284 |
Scribe · Transcriptionwhisper-large-v3-turbo | Transcribes and diarises every client call, then files a searchable summary against the right engagement. | Helios | 412 |
Echo · Speech Synthesisxtts-v2 · cloned brand voice | Gives every synthetic colleague a consistent, brand-owned voice across phone, video and the client portal. | Helios | 967 |
Vega · Video Presenterdiffusion pipeline · avatar-2 | Renders briefing videos and client walkthroughs on demand — one avatar, any language, published in minutes. | Modal | 88 |
Atlas · Analystllama-3.3-70b · vLLM | Reads the warehouse, drafts the weekly client report and flags the three numbers a partner should look at. | Helios | 3,140 |
What changed, measured
Every story below names the situation, the move and the measured outcome. Where a client cannot be named, the sector is named instead.
From a nightly batch to a streaming backbone, without a rewrite
Risk decisions ran on data that was up to twenty-six hours old, because everything downstream waited on a nightly batch that had stopped fitting in the night.
A private inference platform, because the data could not leave
Clinical documents could not be sent to a third-party model, so an entire class of work stayed manual while the organisation watched competitors automate it.
Nine teams, nine platforms, one contract
Nine product teams each ran their own Kubernetes setup, each with its own ingress, secrets and monitoring, and a cluster upgrade had been postponed for two years because nobody knew what would break.
Fixed shapes, so starting is a decision rather than a negotiation
Most engagements begin with a two-week sprint or a five-day diagnostic. Both end with something written that you own, whether or not the work continues.
Catalyst Sprint
2 weeksA written decision and a working prototype for one bounded question.
- A framing session that turns the question into a decidable one
- A working prototype on your own data or your own estate
- A written decision record with the alternatives and the reasoning
- A costed plan for what follows, if anything should
Architecture Diagnostic
5 daysAn independent written assessment, a ranked risk list and a costed roadmap.
- A read of the code and the infrastructure, not only of the documentation
- Interviews with the people who operate it
- The ranked risk register, each item with its blast radius
- A management presentation of the findings
Platform Foundation
8–12 weeksA production platform on your hardware or your cloud, handed over and operable by your team.
- The cluster estate, built declaratively and owned by you
- The platform layer: ingress, storage, secrets, policy, observability
- The delivery path, with promotion and rollback
- Runbooks, a written platform contract and a recorded handover
Tell me what you are trying to build.
A reply from a named person within one business day. If it is not work we should take, I will tell you that, and point you at who should.
Dr.-Ing. Ricardo Yuki Saito · Founder & Chief Executive