Good fit
- AI or agent systems touching spend, customer answers, memory, or ops workflows.
- You can share architecture, traces, prompts, policies, or failure examples under scope.
- You want a diagnostic, one focused fix, or a larger implementation.
What I offer
Three fixed-scope engagements. Prices in USD. Remote. Written deliverables—not open-ended billing. Pick the offer that fits, or describe the problem if you are unsure. Next step: send a short brief.
Entry
Production AI Diagnostic
From $5,000 · 1–2 weeks
A focused review of one production problem. I trace what is going wrong, why it matters, and what I would change. You get written findings, priorities, and a walkthrough.
Delivery
Focused AI Systems Sprint
From $18,000 · 4–8 weeks
Fix one important production AI problem end to end. Typical work includes agent controls, cost enforcement, evals, grounding, workflow reliability, or failure recovery.
Larger scope
Production Implementation
From $40,000 · typically $75,000–$100,000+
Larger systems work spanning multiple workflows, integrations, or longer production ownership. Scope and price confirmed after the problem call.
Something else going wrong?
If you have a recurring production AI problem that doesn’t fit neatly above, send a short description.
These are independent options. Most teams start with a Diagnostic or Sprint. A Sprint can start without a Diagnostic when the problem and owner are already clear.
Reliability Partner
From $3,000/mo after a Diagnostic, Sprint, or Implementation—when continuing reliability work is still useful. One focused track per month, not full-time embedding.
Fit check
Best when the production problem is already real.
Strongest fit: production or near-production system, a technical owner, and enough signal to investigate.
Not a fit
- Idea-stage demos with no owner or production path.
- Open-ended staff augmentation without a defined outcome.
- Guaranteed AI accuracy, compliance certification, or work outside agreed scope.
What you do next
How engagements run
Clear sequence. No obligation until you choose a scoped engagement.
1. Send a problem brief
Short, non-confidential description of the system and what is going wrong. I reply within one business day.
Required first step2. Problem call
30 minutes on what is failing, what you have tried, and what it is costing you.
If there is a fit3. Choose the engagement
Diagnostic, Sprint, Implementation—or I tell you I am not the right fit.
Scoped delivery4. Measure the result
Agree what changed. Continue only if there is continuing value.
OptionalReady to start?
Prefer email for a short intro, or if your company blocks web forms.
Selected proof
Three self-built reference systems demonstrating the patterns used in engagements—not client case studies or live customer deployments.
Why teams trust this work
Top 0.5%
Kaggle Competition Expert · global rank
UZH MSc
Data Science
DevRev
Applied AI · LLM workflows
Zurich
Prior applied AI · AXA XL & Sulzer
Background
I am an independent AI Systems Engineer and founder of KazenAI. I diagnose and fix hard production AI problems—reliability, cost, evaluation, permissions, and agent control.
Previously at DevRev, I shipped LLM systems across live customer workflows. My broader experience includes applied AI work with AXA XL and Sulzer in Zurich, an MSc in Data Science from the University of Zurich, and Kaggle Competition Expert, ranked in the global top 0.5%.
Work is fixed-scope: Diagnostic, Sprint, Implementation, and occasional Reliability Partner tracks. KazenAI is my independent R&D on production AI control and reliability. Selected components may be used where they fit; client systems and confidential data stay isolated under the engagement agreement.
Send a problem brief
Describe the system and what is going wrong—non-confidential only. I reply within one business day. If there is a fit, I send a private link for a 30-minute problem call. This is not a purchase or booking.
Prefer email? [email protected]