Email Machine, concept to production
Co-founded and scaled an email platform from one retailer's internal tool to a SaaS running 100M+ emails/month for 196 brands.
For SME & product teams
Get one AI thing into production: properly
For founders, ops leads, and product managers at 10–250-person companies.
Situations
When teams call me in
- We have a pile of docs nobody can search.
- The team pastes into ChatGPT with no guardrails.
- We want an internal assistant but IT is nervous about the data.
- We built something and can’t tell if it’s any good.
Use cases
By capability
Each capability at the scale of one team: the result to expect, and what has to be in place before it’s worth starting.
RAG & assistants
-
An internal assistant over your docs, wikis, tickets, and contracts.
-
Support deflection over a public help centre with citations back to the source article.
Chatbots
-
A customer-facing FAQ / support bot, lead qualification, or booking flow: scoped, not "answers anything".
Agents & workflow automation
-
Automate a repeatable back-office chain: draft the quote, assemble the report, triage the inbox, with a person approving anything that leaves the building.
Classic ML
-
A model that beats a rule-of-thumb your team relies on: lead scoring, demand and staffing forecasts, anomaly detection.
AI observability & evals
-
Trace-level logging and evals that run in CI, so a regression pages you instead of a customer. Cost and latency on a dashboard; guardrails for PII and prompt injection.
Reality check
Before you start
Note
What has to be true first
Before a build is worth starting, three things have to be true: the data has to be reachable, permissions have to be sorted, and someone on your side has to own the thing after I leave.
It’s an honest filter. If that isn’t in place yet, the intro call says so, and we scope the groundwork first rather than pretend the model is the hard part.
Note
Already shipped something? Start here
If you already have an LLM feature live, the observability and evals pillar is the cheapest high-value way in. You find out what is actually wrong, and how often, before spending on a rebuild.
It’s a fixed scope of a week or two and ends in a written report. You can act on it yourself or bring me back for the fixes.
Approach
No big bang
One use case at a time: scoped, measured, and handed over with the evals to keep watching it. If you want ongoing ownership after that, an optional fractional retainer picks it up. Otherwise the engagement ends and you keep everything.
Track record
Systems I’ve built
Prior in-house work close to this: a product built from concept to production, and the real-time data platforms that back this kind of system.
A real-time Data Management Platform
Product for a DMP processing 500M+ user events/day for 50+ media agencies: real-time segmentation, privacy-first.
Packages
Four ways to start
Fixed scope, fixed fee: quoted after the intro call.
AI Discovery Sprint
For Teams unsure where AI actually pays off.
Scope
- Your situation, your data, your constraints: reviewed
- 2–3 candidate use cases, ranked by impact and effort
- Approach decision per case (LLM API vs a model, RAG vs a recommender, build vs buy)
- Success metrics defined up front
- A one-page architecture sketch
You get
- A prioritised use-case shortlist
- An approach recommendation with the reasoning
- Success metrics and an effort/impact estimate
- The architecture sketch
Outcome You know which AI work to do first, why, and what 'good' looks like.
Fixed fee, quoted after the intro call.
Book the intro callPrototype Build
For A chosen use case that needs to exist and be measured.
Scope
- A working prototype, integrated with your systems
- A baseline evaluation: offline metrics, an A/B or sanity checks
- Documentation and a next-steps roadmap
- A handover to your team
You get
- The integrated prototype
- The evaluation and its numbers
- Docs + roadmap
- A working session to hand it over
Outcome One AI feature in front of real users, with numbers you trust and a team that can run it.
Fixed fee, quoted after the intro call.
Book the intro callAudit & Fix
For An existing RecSys / ML / LLM feature that underperforms or isn't trusted.
Scope
- An audit of the metrics, the data, and the architecture
- A quick-wins list
- An eval harness so regressions get caught
- An improvement roadmap with estimated impact
- For e-shops: an EU AI Act readiness pass (help, not legal advice)
You get
- The audit findings
- The quick-wins list
- The eval harness
- The roadmap
Outcome You know why it underperforms, what to fix first, and how you'll know it improved.
Fixed fee, quoted after the intro call.
Book the intro callFractional AI Lead
For Teams shipping AI without senior AI ownership.
Scope
- Roadmap ownership
- Architecture and PR review
- Eval governance: keeping the numbers honest
- Vendor and model decisions
- Mentoring the team
You get
- A maintained roadmap
- Regular reviews
- Eval and cost/latency oversight
- A team that levels up
Outcome Senior AI judgement in the room every week, without a full-time hire.
Monthly retainer, min. 3 months: quoted after the intro call.
Book the intro callFAQ
Common questions
We’re worried about our data leaving the building: how does that work?
It doesn’t have to leave. The model call is the only thing that goes out, and that can run through an EU region or an API that contractually doesn’t train on your inputs. Retrieval, logs, and your source documents stay on infrastructure you control. The data flow is written down before any build starts, so IT can see exactly what crosses the boundary.
Do we need a data scientist on our side?
No. You need one person who owns the use case and can answer questions about the business: usually an ops lead or a product manager, a few hours a week. I bring the modelling, evaluation, and engineering. If you already have a data person they speed things up, but the work is built to hand over to a generalist.
How small is too small for this?
Below about 10 people it’s usually cheaper to use off-the-shelf tools and wait. The work pays off when you have a repeated, documented process a general assistant gets wrong, or a pile of documents people lose real hours searching. If you can’t point to a specific task and roughly what it costs in time, the intro call will say so rather than sell you a project.
Can we start with just an eval / audit?
Yes, and if you already have an LLM feature live it’s the right place to start. A week or two of observability and evals tells you what’s actually failing and how often, before you spend on a rebuild. It’s a fixed scope, it stands on its own, and it ends in a written report you can act on yourself.
What does the fractional retainer actually cover?
A set number of days a month owning your AI work: roadmap, reviews, vendor calls, and keeping the evals honest as things change. It’s for teams that have one thing in production and don’t want to hire a full-time lead yet. Month to month, cancel with notice; the point is that someone senior stays accountable.
Tell me where your AI is stuck.
Thirty minutes, no pitch. Your situation, your data, whether this is the right problem.