SEISMOGRAPH
Aarhus, DK · LLM Drift Defense
Free scan
Early-warning for silent LLM provider drift

Your LLM didn't get worse.
It changed — and nobody told you.

Providers ship silent model changes with no version bump and no changelog. Same 200 OK, same latency — but JSON fidelity slips and outputs drift. Standard monitoring is blind to it. I build the early-warning layer that isn't.

Request a free Drift Exposure Scan See the live dashboard →

Open-source engine · Apache-2.0 · privacy-preserving by design · you own every result.

For teams shipping on OpenAI, Anthropic, Mistral or any third-party LLM API — where a silent model change means broken JSON, degraded answers, and support tickets before you know why.

Proof of process

Not a claim. A receipt.

Two-person teams pitch with a deck. I pitch with a reproducible, seeded backtest you can re-run yourself.

38 days
Flagged the Claude Sonnet 4 silent degradation before the provider's official postmortem — reproducible backtest, during the 0.8% misrouting window, before user-visible symptoms.
4 models
Live public “model weather” dashboard, refreshed continuously. No login — open it right now.
127 tests
Green test suite, change-point (CUSUM) detection, audit-grade evidence export. Engine is open-source and auditable.

In the backtest, detection occurred during the 0.8% misrouting window — 38 days before the official postmortem and 19 days before the escalation became visible to users. The methodology is hash-committed, so the next call is a prediction, not a postmortem.

Evidence: live dashboard · open-source engine · dev.to writeup · pip install seismograph-probe

Your path, start to protected

Three steps. First one costs nothing.

Step 1 · today

Send your model stack

One email listing the LLM APIs you depend on. No call, no system access, no NDA needed.

Step 2 · ~48 hours

Get your exposure map

Your top-3 silent-drift exposures, ranked, with the exact metrics that would move first. Yours to keep either way.

Step 3 · ~1 week

Probes go live

If the map warrants it: canary probes on your real model tuples, tuned thresholds, alerts wired into Slack or PagerDuty. Fixed price, reviewable deliverable.

How we work

Start free. Scale only if it's worth it.

Flat pricing, concrete deliverables, fixed timelines. No open-ended “let's chat” retainers to start.

Start here

Drift Exposure Scan

Free

~48 hours · no system access

  • Send me your model stack (which APIs you depend on)
  • I map where a silent change would hit you first
  • You get your top-3 drift exposures + the metrics to watch

Zero risk. You leave with a watchlist worth more than the time it took.

Most popular

Drift Baseline

€2,500 flat

~1 week · reviewable deliverable

  • Canary probes stood up on your real model tuples
  • Behavioral baselines + CUSUM thresholds tuned to your traffic
  • Written exposure report + a concrete alerting plan (Slack/PagerDuty/webhook)
  • You own everything; nothing proprietary leaves your perimeter
Book a Baseline
Ongoing

Drift Desk

€3–6k / month

standing early-warning line

  • Continuous watch across your model dependencies
  • A heads-up the moment a provider shifts — plus a monthly “model weather” briefing
  • Priority response when a provider incident hits

Cheaper than one bad week in production.

Need it built into your stack and handed over working? A full Proof-of-Process Build (in-VPC probe fleet wired to your alerting, runbook, handover) is scoped per engagement — ask in your scan.

Built for teams that can't afford a silent break

Privacy-preserving by design.

No raw prompts or outputs. Ever.

Only derived, non-reversible features cross the boundary — token counts, a JSON-valid flag, SHA-256 hashes. Raw text is never stored or transmitted.

Differential privacy on every metric.

Numeric signals carry Laplace DP noise with a tracked epsilon budget. Aggregated, anonymised, defensible.

Runs in your VPC.

Probes deploy inside your perimeter. Your data never depends on my infrastructure. Audit-grade evidence export for compliance.

Structurally neutral.

I don't sell you a model, so I have no reason to tell you it's fine. The engine is Apache-2.0 and auditable line by line.

Every week unmonitored is a week exposed.

The Anthropic routing bug degraded output for weeks at 0.8% of traffic while every dashboard stayed green — a seeded backtest flags it 38 days before the postmortem. Your free scan takes one email and ~48 hours.

FAQ

Straight answers.

How is this different from Datadog / LangSmith / eval tools?

App-level observability tells you your outputs got worse. It can't tell you whether the model itself changed — and it's often channel-conflicted (it sells to the providers it would have to indict). SEISMOGRAPH watches the provider side, neutrally, across time.

What do I actually get?

The free scan: your top-3 drift exposures within ~48h. The Baseline: live canary probes, tuned thresholds, a written report, and an alerting plan — a reviewable deliverable in about a week, not an open-ended retainer.

Is my data safe?

Raw prompts and outputs never leave your perimeter — only DP-noised aggregate features. Probes run in your VPC. You own every artifact. Full technical detail is in the open-source repo.

Why should I trust the detection?

Because it's falsifiable. The 38-day Claude Sonnet 4 call was made on a seeded, reproducible backtest you can re-run. The methodology is hash-committed up front.

What isn't ready yet?

Honest answer: the public network effect (cross-org quorum) compounds with adoption and is early. Today the value is the in-VPC early-warning layer and the neutral monitoring — which stand on their own. I'd rather tell you that than oversell it.

Built by
TR

Tetiana Radchenko

AI infrastructure / backend engineer based in Aarhus, Denmark. I built SEISMOGRAPH from scratch to solve the most under-monitored failure mode in production AI: silent, provider-side model drift. Every engagement ships under my personal sign-off.

One free scan tells you if it's worth your time.

Send me the model APIs you depend on. Within ~48 hours you get your top-3 silent-drift exposures and the metrics that move first — plus a fixed quote if you want the Baseline.

or email me directly: tatyan.radchenko@gmail.com