Theremia

Market position

Early-stage venture-backed precision-medicine company positioned between AI drug development, causal inference in healthcare and real-world evidence. It differentiates by focusing specifically on CNS indications and on interindividual pathophysiological variation rather than single biomarkers or genomic subgroups. Investors cited its algorithmic approach rooted in real-world data as complementary to traditional theoretical and clinical methods. Sources name no direct competitors.

Sources emphasise: a focus on modelling interindividual pathophysiological differences to tailor dose, frequency and formulation for population subgroups rather than biomarker- or genomics-based stratification; stated unique access to over one million high-quality granular longitudinal patient records; an interdisciplinary team spanning chemistry, molecular pharmacology, modelling, physiopathology and neuroscience; close collaboration with practitioners specialised in each pathology so that algorithms capture disease determinants; and dual applicability to clinical-trial design and to repurposing/lifecycle extension of existing assets.

Technology

AI and machine-learning models combined with multi-scale pharmacological modelling that link phenotypic and molecular scales. Capabilities described include detection of responder and non-responder patient subgroups, prediction of adverse reactions using demographic and physiological covariates, and optimisation of pharmacological parameters (dosage, frequency, galenic form, route of administration, drug combinations). The company also describes causal-inference-oriented outcome methodology work around the Win Ratio for heterogeneous populations. Inputs include a stated corpus of over one million granular longitudinal individual patient records and real-world data; models are described as validated with practitioners across prospective cohorts. The public website is built on Framer and hosted on AWS with Cloudflare DNS.

Go-to-market

Direct partnership with biotech and pharmaceutical developers, invited via a 'Book a Demo' / 'Partner with us' motion on its website, complemented by public-private partnerships with medical institutes and clinicians across CNS pathologies. Visibility is built through industry conferences and ecosystem programmes, including a booth at the BIO-International Convention 2026 in San Diego with Business France and France Techbio, and a poster at the 2026 World Parkinson Congress in Phoenix on automated levodopa dose individualization.

Pharmaceutical and biotech drug developers working on central nervous system indications, along with clinical teams and academic/medical institutes involved in validation partnerships. The end beneficiaries described are patients with neurological and psychiatric conditions including Parkinson's disease, depression, Alzheimer's disease and multiple sclerosis.

Geography

Headquartered in the Paris region of France; one directory lists an address in Levallois-Perret (Hauts-de-Seine, Île-de-France). Ecosystem support and affiliations cited include Bpifrance, Agoranov, Station F, Medicen, Entrepreneur First, Better Angle, the Female Founders Fellowship, Deep Tech Pioneers, and the DayOne incubator in Switzerland. Planned commercial outreach includes events in the United States (San Diego and Phoenix) in 2026.

History

Founded in 2024 by Iris Maréchal and Chloé Geoffroy (one directory lists a 2023 creation date and one profile shows Maréchal as Founder in Residence at Entrepreneur First from October 2023 and co-founder/CEO from January 2024). The company announced a €3 million seed round on 7 November 2024 led by Eurazeo and Salica Investments with participation from Entrepreneur First, Tiny VC and Better Angle plus angel investors, intended to expand its team of nine scientists and strengthen its algorithms through medical-institute partnerships. It began with work on Alzheimer's and Parkinson's disease and subsequently extended its validated algorithms to major depressive disorder, receiving the Innovation and Mental Health Trophy in March 2025.

Risks & controversies

Public sources contain conflicting basic data: founding year is given as 2024 in most sources but 2023 in two directories, and one directory lists an implausible headcount band of 201-500 employees while another lists 2-10 and the funding announcement states around ten. Investor lists also differ across sources (Eurazeo/Salica/Entrepreneur First/Tiny VC/Better Angle in the announcement versus Entrepreneur First/Hambro Perks/Motier Ventures and named angels in one aggregator). No clinical validation results, peer-reviewed outcomes, customer contracts or regulatory approvals are disclosed. The company's own content acknowledges obstacles to AI in mental health, including sparse real-world data on 15-28 year-olds, fragmented and inconsistent datasets, limited visibility of AI impact to clinicians, and limited cross-disciplinary collaboration.