The drug is the patient. The schedule is the therapy.

Meridix is the centre-side control layer for cell and gene therapy — readiness scoring, continuous slot optimisation, and one chain of identity across every manufacturer portal.

Measure your attrition See how it works

Autologous CAR-T & TCR-T Batch size one Multi-manufacturer by design

Meridix Console — Cell Therapy Programme, Site 04

Live

CTC

R-4471 is booked into a manufacturing slot on 14 Oct. Is she going to make it — and if she isn't, who should have the slot?

@readiness

Scores probability of being apheresis-ready on the candidate date from labs, disease trajectory, prior therapy exposure and authorisation state. P(ready · 14 Oct) = 0.41 · calibrated on 2,318 prior journeys

@exposure

Checks the therapy history for agents that compromise T-cell yield before collection. FLAG · bendamustine at −9 wks · lymphodepleting · yield risk high

@authorisation

Tracks payer approval and predicts turnaround against the apheresis date, not the infusion date. Prior auth submitted 22 Sep · predicted decision 09–16 Oct

@slotboard

Re-solves the whole site pipeline against open slots across all four products, and decides what to release and when. 37 patients · 11 slots · 4 sponsors · re-solved 14s ago

Recommend release. R-4471 scores 0.41 for 14 Oct; the binding constraint is T-cell fitness, not authorisation, so waiting does not help. Rebook 04 Nov (0.86). Slot released to the reallocation pool — R-4392 at Site 07 scores 0.88 and is ready now. Clinician confirmation required.

Ask about a patient, a slot, or the pipeline…

Readiness Slot board Attrition Chain of identity

The problem

Patients wait for slots that do not exist, and slots go unused because the patient booked into them was not ready — and nobody counts what is lost in between.

≈31 days median vein-to-vein, cross-product
Up to three weeks waiting for a manufacturing slot
6–12 months to activate a single treatment centre
Pre-infusion attrition — unpublished, uncounted, unowned

How it works

One record, one board,
one number per patient.

The coordinator keeps working the way the ward works. Meridix absorbs the fragmentation underneath — four portals, five organisations, and a slot that has to be booked before anyone can know whether it can be used.

Walk through the console

Readiness · R-4471Review

0.41P(READY · 14 OCT)

T-cell fitness−0.29 binding

Authorisation−0.06

Disease trajectory−0.11

Apheresis capacity+0.05

Same patient · 04 Nov0.86

Score readiness, not status

Not a green tick on a spreadsheet. A calibrated probability that this patient will actually be collectable on this date, decomposed into the drivers — so a clinician can argue with the reasoning instead of a number.

Slot board · week 41Re-solved 14s

MON 06TUE 07WED 08 THU 09FRI 10SAT 11

Slots held11

At risk (P<0.5)3

Released this month4

Reclaimed by network3

Solve the board continuously

Match the pipeline against every open slot and re-solve as patients change. The harder question is the one nobody asks today: which slot should be released, and how early, so that somebody else can still use it.

Chain of identityReconciled

09:04COI-8842 received · sponsor portal A

09:04Matched to R-4471 · DOB + MRN + COI hash

11:20Dry shipper LN-2217 · −167 °C · in transit

11:20Excursion window: none

14:41Portal C label mismatch · flagged

14:52Resolved by coordinator · audit written

Portals unified4 / 4

Reconcile the custody record

One patient record across every sponsor portal, with chain-of-identity verification reconciled rather than re-keyed. We integrate with the manufacturer systems; we do not pretend to replace them.

What we believe

Meridix never allocates a slot to a patient. It surfaces readiness, constraints and consequence. Clinicians allocate.

Capabilities

The things no party
with a product to sell can build.

Cross-product selection support, slot reallocation across sponsors, and honest attrition measurement are only available to someone with nothing to sell in the decision. That is an architectural position, not a marketing one.

Read the neutrality argument

The denominator · 24 monthsRetrospective

Intended for therapy412

Reached apheresis301

Reached infusion247

−58Lost to disease progression

−31Lost to authorisation delay

−22Lost to forfeited slot

−54Lost after collection

Operational, not clinical53 of 165

The number nobody has

Registries count the patients who were infused. The losses happen upstream, in fragments, across systems that never meet. We reconstruct the full path from intent to infusion and tell you where your programme actually loses people.

Selection supportNeutral

Third-line DLBCL, prior bendamustine, payer B. What are the realistic options?

Three products are available to this site. Given the exposure history, expected yield favours two of them. Realistic vein-to-vein at your centre: 24d / 29d / 33d, using your own historical medians rather than label figures.

Sponsor influence on rankingNone · not for sale

Product selection support

Which available product fits this patient, and what is the realistic vein-to-vein time for each — measured at your centre. No manufacturer will ever ship this, because it sometimes recommends a competitor.

Network & integrationsLive

Sponsor ASponsor BSponsor C Sponsor DEHR / HL7Cryo courier

Slot utilisation, network+11 pts

Forfeitures reallocated63%

Median release lead time9 days

Capacity intelligence, both ways

Centres see every open slot they could reach. Manufacturers see real demand, true utilisation and why forfeitures happen — the mirror image of their own portal, which today shows them a quarter of the picture.

What you get

Four things that compound,
and cannot be copied by an incumbent.

Structural neutrality

We have no product in the decision. Every incumbent does. Copying us would mean dissolving themselves, which is why the position holds.

The readiness ledger

Who made it, who did not, at which step, and what the leading indicators were — across centres and across products. No registry holds this. It compounds with every journey.

The integration surface

Every sponsor portal, protocol, slot rule and custody convention, maintained as they change. Unglamorous, and a genuine barrier to anyone starting later.

The coordinator's desk

The cell therapy coordinator is the person this exists to serve. Once the day runs in one system instead of six, the shortage stops setting the ceiling on how many patients a centre can treat.

Founder

MGK

Mari Ganesh Kumar Founder

A decade of building models on biological signals, pointed at the one number this field cannot compute.

Mari holds a PhD and MS in Computer Science and Engineering from IIT Madras, where his thesis applied machine learning and signal processing to bio-signals — speech, EEG and neuronal recordings — published with Hema A. Murthy and Mriganka Sur, and recognised with the Institute Research Award in 2021. The work was consistently about the same thing: extracting a reliable, decomposable signal from noisy biological data where the ground truth is observed and the stakes of a wrong call are real.

He spent three years as an Applied Scientist at Amazon working on large-scale speech and language systems, and is currently a Senior Software Engineer in AI/ML at Google, after earlier engineering roles at MTX Group and Zoho. He was awarded the medal for best outgoing student at Thiagarajar College of Engineering in 2015.

Readiness is a prediction problem with real labels — the outcome is observed every single time. What was missing was never the model. It was anybody assembling the data across the organisational boundary.

founder@meridix.cloud LinkedIn

Doctorate & Master's: Computer Science & Engineering, IIT Madras
Research focus: Applied ML and signal processing for bio-signals
Industry: Google · Amazon · MTX Group · Zoho
Recognition: Institute Research Award, IIT Madras, 2021

Trust & data

Governance is architecture here,
not a policy document.

The centre owns its data. Anything that moves to the network layer is de-identified, and no cross-centre model training happens without a contractual opt-in. What sponsors can see and what selection support uses are separated in the system, not in a promise — because the moment neutrality is compromised, it stops being an asset and becomes the liability.

Readiness is explanatory by construction: every score decomposes into its drivers. And the model is audited for whether it disadvantages patients by payer type, distance, language or any proxy for income — with the methodology published and reviewed externally, not kept as an internal policy.

HIPAA · BAA GDPR SOC 2 Type II 21 CFR Part 11 Centre-controlled data Tenant isolation RBAC + full audit trail De-identified network layer No training without opt-in Published fairness audit Clinical & ethics board

Meridix is at design-partner stage. The controls above are the committed architecture and the certification roadmap — we will tell you exactly which are audited today on the first call, rather than implying all of them are.

Data residencyPer-centre, in region

Model trainingOpt-in, contractual

Score explainabilityEvery score, every driver

Fairness auditPublished, external review

Allocation authorityClinician, always

Book a demo Talk to the founder

Get started

Let us measure your programme's attrition.
No integration. No commitment.

Request the analysis Email the founder

Take one programme's last two years. We reconstruct every patient intended for therapy and trace where each one went. Most centres have never seen this analysis of themselves.