Review of technical decisions
Architecture, models, data handling: we assess decisions before they become expensive to reverse.
Your technical reference point
AI consultant
Sometimes you don't need to hand over a project — you need someone to think with before deciding. A senior engineer who knows your context and is consistently available, without hiring a full-time role.
30 minutes with the person who would actually work with you, not a salesperson.
- A stable technical counterpart
- No vendor ties
- Commitment that flexes over time
- Skills transferred to your team
When it helps
When a consultant beats a project
A consulting arrangement makes sense when what you need isn't a deliverable but recurring technical judgement.
- AI decisions are made in-house, but nobody can assess their technical consequences.
- Your development team is capable, but new to AI, and needs someone to steer around the known mistakes.
- You need to evaluate vendors, quotes or candidates and want an opinion with no stake in the answer.
- You need technical direction on AI, but not enough work to justify a hire.
If you already have a defined project to deliver, a development engagement is more efficient than ongoing consulting.
How we support you
What an AI consultant actually does
Not generic consulting hours: recognisable interventions, each with an outcome.
Vendor and proposal evaluation
We read quotes with a technical eye: what's genuinely included, which running costs surface later, where lock-in begins.
Support for your internal team
Code review, working sessions, answers when the team is stuck. The goal is to be needed less over time.
Hiring support
Defining profiles and running technical interviews, to tell who can really build AI systems from who has only read about them.
Part-time technical leadership
A regular presence in product and platform decisions, with the continuity of an internal role at the cost of an external one.
How it works
How the relationship is built
A consultant is only useful with context — the first sessions exist to build it.
Understand the context
Current situation, pending decisions, skills already in the company. We also check that ongoing support is genuinely the right format.
Technical immersion
We go through your systems, data and existing code. Without this, opinions stay generic and are worth little.
Regular presence
Periodic sessions plus availability in between, for the decisions that don't wait for the calendar.
Growing autonomy
As the team gains experience, the commitment shrinks. A good consultant works to become less necessary.
What you can rely on
The areas we give an opinion on
Experience built by shipping systems we then had to maintain — the part that teaches the most.
Model selection
Commercial or open, large or small, fine-tuned or not: the choice changes running costs and data constraints.
RAG and agent architectures
How to structure retrieval and system autonomy without building something ungovernable.
Evaluation and quality
How to measure whether an AI system actually works, instead of trusting a handful of examples that went well in a demo.
Running costs
What drives the monthly cost of an AI system, and which design choices make it grow unexpectedly.
MLOps and production
Release, monitoring and model updates: the distance between a prototype and a system that survives daily use.
Compliance and governance
GDPR and the EU AI Act as design constraints, including system classification and the responsibilities that follow.
Working together
How an engagement is structured
The format depends on how continuous the need is and how much team there is to support.
Periodic support
Regular sessions plus availability in between, to accompany decisions while they're being made.
When an internal team is building.
Part-time technical leadership
A reference role for platform and product choices, with explicit accountability for the recommendations.
When there's no senior AI voice and hiring one is premature.
Single-decision opinion
A bounded intervention: assess a proposal, settle an architectural doubt, give a second opinion on a project.
When there's one question, but it carries weight.
We don't publish hourly rates because the commitment varies widely: session frequency, depth of the technical immersion, duration. On the first call we work out how much presence is really needed — usually less than expected — and the proposal follows from that.
FAQ
What people ask before starting
The Babel Peak engineers who build the systems, starting with Michelangelo Bagnara, CTO and co-founder. There's no handover between whoever sells and whoever works: the person on the first call is the person who stays.
Project consulting has a defined scope and closes with documents. Here you're buying continuity: someone who knows your context and can think it through with you, including on small questions. If you need a single answer, the project format costs less.
Yes, and that's the most effective mode. Code review, joint working sessions and decisions taken together: the team grows and dependence on us falls, which is the desirable outcome.
Yes, and we hold no reseller agreements with anyone. We assess proposals on technical merit and total cost over time. If the best choice is a third-party vendor, that's what we'll tell you.
It usually starts with a defined commitment for the first months — enough to absorb the context and accompany the first decisions. Then it adjusts: some continue for a long time with a light presence, others close once the team is autonomous.
Related services
If you need a defined project rather than ongoing support, these are the right paths.
AI consulting
We help you decide what's worth doing with AI, what isn't, and in what order. Then, if it makes sense, we build it: we're the same people who write the code, so the advice has to survive contact with reality.
AI for business
You don't need to reinvent the company to use AI. We start from a process that costs you hours every week, test it on real data within weeks, and scale only once the numbers justify it.
AI software development
Most AI projects stop at the demo. The distance between a prototype that works in a meeting and a system that survives daily use is made of integrations, error handling and monitoring — that's the part we do.
Let's talk
Let's see whether you actually need a consultant
Half an hour to understand the context and how much presence would help. If a one-off intervention is enough, we'll say so.