Mantis Biotech's Parva Uses AI to Rethink How Women's Cognitive Health Is Diagnosed

Mantis Biotech (搜索)搜索该词条, a startup led by Georgia Witchel, has launched Parva (搜索)搜索该药品, an AI-powered consumer software designed to close the gap between what patients sense about their own cognitive health and what medicine can currently measure. The product quietly reads how a person types and clicks, then flags cognitive change against that person's own history rather than against population averages—a shift Witchel describes as "you against yourself" rather than "you against the population."

A Diagnostic Model Built on the Wrong Baseline

Cognitive care, Witchel argues, runs on a schedule that works against the biology it is meant to track. Treatment for conditions "like long COVID (搜索)搜索该疾病, dementia (搜索)搜索该疾病, perimenopausal syndrome (搜索)搜索该疾病 is completely based on in-person appointments that will happen once a month at relatively random intervals." A single monthly visit captures a mind at one arbitrary moment, then treats it as a trend.

For women, that design flaw compounds. Symptoms shift with hormones, and as Witchel notes, "your emotions, your ability to self-regulate, is going to fluctuate with your hormones, with your monthly cycle." The stakes are substantial: a 45-year-old woman carries a 1 in 5 lifetime risk of Alzheimer's (搜索)搜索该疾病, double the 1 in 10 a man faces. Women were also found 31% more likely to develop long COVID (搜索)搜索该疾病 than men in the federal RECOVER study.

The root of the problem, according to Mantis Biotech (搜索)搜索该词条, is historical bias in medical research. When studies predominantly feature male subjects, the "normal" physiological and cognitive baselines established often reflect male biology. Conditions such as ADHD (搜索)搜索该疾病, which frequently presents in women as inattentiveness, disorganization, or internal restlessness rather than the classic hyperactive male presentation, are consequently diagnosed much later—if at all. The same pattern holds for dementia (搜索)搜索该疾病 and perimenopausal cognitive changes, which are often dismissed as "normal aging" or "brain fog."

A Baseline of One

Parva (搜索)搜索该药品 throws the population reference frame out entirely. The software installs on a laptop and reads passive signals—keystroke latency, mouse movement, how long attention lingers on a tab, how quickly someone answers a loved one—then stores them in a HIPAA-compliant database and weighs each day against the user's own earlier readings.

The shift reframes diagnosis itself. Taking ADHD (搜索)搜索该疾病 as an example, Witchel says Parva (搜索)搜索该药品 asks whether a woman improved against her own pattern, "rather than... the traditional method, which is just, do they fit into a traditional male pool as diagnosable as ADHD."

Witchel's technical argument is that pattern-matching AI flattens the very people who most need to be seen. "Statistical AI really breaks down at the individual level. It's really, really good at predicting populations," she argues. Her analogy: ask an image model for a face, and "it'll give you a perfect human face. AI is very, very bad at doing uneven faces."

From Injury Prediction to Cognitive Care

Witchel did not begin in neurology. She built physics simulations to predict when bodies break, work that drew her into professional sports because, as she puts it, "the number one buyer of a physical simulation of a human is someone who really, really cares if a human is going to get injured, and that is a pro athlete."

She then followed the unmet need to a larger market. The total economic burden of dementia (搜索)搜索该疾病 reached roughly $781 billion in 2025, while depression costs U.S. employers an estimated $187.8 billion a year.

Funding and Market Position

Mantis Biotech (搜索)搜索该词条 has raised $7.4 million in funding, led by Decibel (搜索)搜索该词条, with Y Combinator (搜索)搜索该词条 and Liquid 2 (搜索)搜索该词条 participating. Jon Sakoda of Decibel, who led the round, observes that "the next frontier will be models that understand each person as an individual. Mantis shows what's possible."

The company straddles three young markets whose edges are still soft: digital biomarkers, forecast to reach $17.73 billion by 2031, according to Markets and Markets; behavioral biometrics, projected at $11.38 billion by 2031, according to Mordor Intelligence; and healthcare digital twins, headed toward $3.55 billion by 2030, according to AppInventiv.

Its nearest rivals cluster in digital biomarkers, among them Mindstrong, Altoida, and Evidation Health, though most chase a single signal or a single disease. For scale, the clinical digital-twin company Unlearn AI has raised more than $130 million, more than seventeen times Mantis's seed.

Access and Clinical Integration

Access to Parva (搜索)搜索该药品 runs through clinics that manage long-term cognitive conditions rather than a pharmacy shelf. Witchel's team places Parva with ADHD (搜索)搜索该疾病 and executive-function clinics, outpatient psychiatry practices, and therapy practices treating depression, burnout, ADHD, or neurodivergence. Perimenopause clinics and neurology practices round out the current channels, matching the software to the specialists patients already see.

Parva (搜索)搜索该药品 is not a cure and, legally, it is not yet a diagnosis. Mantis flags risk levels and defers the verdict, telling users "this is the chance" rather than handing down a sentence—a boundary Witchel draws on purpose while the company works toward FDA clearance and reimbursement.

A Patient's Perspective

Robyn Schlicher, who spent years trying to measure her own cognitive decline after a viral illness in 2020 left her with brain fog, memory gaps, and trouble reading, signed up on the spot. "Parva (搜索)搜索该药品 determined my baseline upon joining, and now I let it work in the background as I work, with no need for wearable or invasive devices," she explains.

For Schlicher, whose decline was once visible only to herself, a continuous record of who she was yesterday changes the conversation with every doctor she sees. "I consider it a priceless gift for my overall health and ability to run my business," she says.

The larger promise, as Witchel frames it, is harder to price. If a baseline of one can catch what population averages miss, the people medicine has historically measured last may finally get measured first.