# QCPlanner — QC laboratory planning

> Sample intake from LIMS, then a constraint solver that schedules the laboratory against real capacity, analyst competency and predicted OOS risk.

- Where it stands: Pilot
- Quality · Synlogica
- Canonical: https://synlogica.ai/applications/qcplanner/
- Language: en-US

QCPlanner plans the QC laboratory: sample registration and LIMS intake, then a constraint solver that schedules work against real instrument capacity, analyst competency and predicted out-of-specification risk. The solver decides; the models only steer it.

## The problem

A QC laboratory is scheduled on a whiteboard and in the head of the supervisor. Instrument time, analyst qualification, method duration and release deadlines all compete, and the plan is rebuilt by hand every time a sample arrives late or an assay fails. The cost is release lead time nobody can attribute to anything.

## What it does

1. Takes samples in from LIMS extracts and registers them with method, priority and release deadline.
2. Schedules the laboratory with a constraint solver against instrument capacity, analyst competency and method duration, and re-plans when reality changes.
3. Feeds the solver with model predictions — OOS risk, expected duration — while keeping the decision deterministic and every plan version on the audit record.

## Key capabilities

- **Sample registration and LIMS intake** — Manual registration, daily LIMS extracts (CSV/Excel) with per-tenant column mapping and order-independent duplicate detection.
- **Constraint solver that schedules the laboratory** — A CP-SAT optimiser assigns tests to instruments and analysts against real capacity, competency-gated assignment, deadline windows and an OOS-risk objective — and re-plans when reality changes.
- **ML that steers, never decides** — Duration, OOS-risk and throughput predictions feed the solver and are retrained on the tenant's own history; the deterministic solver remains the decision core.
- **A local assistant with gated execution** — The assistant runs on a local model, so batch data never leaves the server; it answers freely but every change is proposed, confirmed by a person and audited.
- **Result lifecycle and laboratory release** — Enforced test and series state machines, OOS → retest workflow and an e-signed laboratory disposition behind a readiness gate.

## Who it is for

QC laboratory managers and supervisors in pharmaceutical manufacturing and contract laboratories who plan against instrument capacity and release deadlines every day. Currently running as a local deployment with early users.

## Where it stands

Pilot — in use with early customers. Scope, data and success criteria are agreed per pilot, and we say what is production and what is prototype before you ask.

## The same foundation as every Synlogica application

Every application in the portfolio carries the same controls, so an inspector who has seen one knows how to read the next.

- Versioned records — a record can be superseded, never silently altered.
- Electronic signature on every controlled step.
- An audit trail an inspector can follow, hash-chained end to end.
- Runs on your own tenant, hosted in the EU.

## Read more

- [Probabilistic decisioning for pharma — Monte Carlo, Bayesian and Arrhenius](https://synlogica.ai/resources/probabilistic-decisioning/)
- [Accelerated stability models compared — Arrhenius, ASAP, AccelStab & probabilistic decisioning](https://synlogica.ai/resources/stability-models-compared/)

## Talk to us

Bring a real case — a recent excursion, an invoice you doubted, a procedure that just changed — and we walk it through together. If it is not a problem for software, we will say so.

- Book a 30-minute call: https://calendly.com/adam-karpinski-synlogica/30min
- Email: adam.karpinski@synlogica.ai
- Privacy Policy: https://synlogica.ai/privacy/