Specimen receiving

Most clinical lab errors happen before the analyzer.

Kinesis Labs builds a benchtop station that takes a sealed courier bag through unbagging, tube inspection, LIS accessioning, sorting and centrifugation — and hands finished racks to staff the same way at 3 a.m. as at 3 p.m.

On the bench

The prototype, working real tubes.

Each task runs in simulation first, then on the bench: the arm reads each specimen, decides what happens to it, and checks the accepted ones into the LIS. The bench clip plays muted — unmute for the audio.

  • Reads the requirementReasons through what the ordered test needs, and takes only a specimen that satisfies it.
  • Holds the bad onesDetects error specimens and sets them aside instead of passing them downstream.
  • Decides what gets spunKnows which specimens go to the centrifuge, and runs it.
  • Accessions in the LISChecks each accepted specimen in directly, so the order and the physical tube are linked.

01Simulation

A simulated copy of the bench, where each task runs before it goes to the hardware.

02Physical

Bench prototype, filmed August 2026. A working system on a bench, not a station deployed in a lab.

The problem, in numbers

The receiving bench is where the errors start and the hours go.

46–68%

of laboratory errors happen before testing ever starts.

Most lab error is pre-analytical: the wrong tube for the order, a specimen that was never received into the LIS, a tube spun that should not have been. The analyzer is rarely the problem. The bench that creates the problem is the one still run entirely by hand.

Published error-distribution literature
$190–225k

a year to keep one receiving desk staffed around the clock.

Specimens arrive on the courier's schedule, so the desk is covered on three shifts. That is the fully loaded cost, every year, for work that is mostly unbagging, checking and racking.

BLS wage data, fully loaded
7–11%

of clinical lab positions sit vacant, higher in some regions.

The person you need at 2 a.m. is the hardest one to hire and keep. When the seat is open, receiving gets covered by whoever is available, and that is exactly when the same tube starts being handled two different ways.

Clinical laboratory workforce vacancy data

Put those together and the picture is simple. The step where most laboratory error begins is the step that is staffed around the clock by the role that is hardest to fill, and it is the one step a lab cannot slow down, because the courier does not wait.

How it works

Sealed bag in. Racked, accessioned, spun specimens out.

One station, six steps, on the receiving desk you already have.

  1. Unbag

    Opens the sealed courier bag and removes the contents, including the requisition paperwork, without disturbing the tubes.

  2. Inspect

    Reads each tube: label, tube type and additive, fill volume, and physical integrity. Flags any conflict between the ordered test and the tube it arrived in.

  3. Accession into the LIS

    Checks each accepted specimen into the laboratory information system so the order and the physical tube are linked before anything moves.

  4. Sort into racks

    Sorts into spin, no-spin and hold racks based on the ordered test and the tube it arrived in.

  5. Load, run and unload the centrifuge

    Balances and loads the spin rack, runs the correct protocol, and unloads it when the cycle finishes.

  6. Signal staff

    Tells staff which racks are ready for the analyzers and which specimens are being held for a person to resolve.

Held specimens are deliberately not accessioned. A tube in the hold rack has not been checked in, so nothing downstream can pull it, bill it, or treat it as received before a person has ruled on it.

Accessioned specimens move from “collected” to “in-lab” in the LIS, which is the state change everyone downstream is actually waiting on.

Physical AI

Trained by teleoperation. Deployed autonomously.

Teleoperation is how the system learns, not how it runs.

Learns from the bench

A licensed technologist drives the arms through real receiving work. Those demonstrations are the training data.

Runs on its own

What ships is a trained policy running on the station in your lab. No remote operator, no per-specimen human in the loop.

Judgment at the sealed bag

The hard part is not moving tubes. It is deciding, tube by tube, what gets accessioned, what gets spun, and what gets held.

Two arms, one bench

A benchtop station that works the receiving desk you already have, not a track line that replaces the room.

This is the single most common misreading of what we do, so we state it plainly: the customer receives a trained policy, not a remote operator. Nobody is driving the arms from off site.

See it on your workflow

Hand the receiving desk to the station.

Tell us what arrives at your bench, or pick a time to talk it through with Alina.

Get in touch

Who we are

Alina Minikhanova

Alina Minikhanova

Co-founder · Knows the task

Clinical Lab Scientist

  • 10 years doing this exact work by hand, across every department of a hospital lab.
  • 3 years building the software systems labs use to track samples.
  • Labels the training data, judges whether a demonstration was good, and can say why a robot got it wrong.
  • Knows, by name, the lab directors this station is built for.
Armin Foroughi

Armin Foroughi

Co-founder · Builds the system

Robotics · Electronics · Machine learning

  • Has already automated a lab assay end to end with a robot arm.
  • Has built several drones and robots, from the electronics up.
  • Computer vision and robot learning; trains models on GPU clusters.
  • Biochemistry background and real wet-lab time, so the constraints need no translation.
  • MS Data Science, University of Virginia. BSc Biochemistry with Data Science, UC San Diego.