# Semalytix

> Semalytix builds Project Atlas, an agentic AI platform that turns the global, multilingual patient voice (Patient Experience Data, or PED) into continuous intelligence for medical, commercial and safety teams in pharma.

## Core pages

- [Home](https://www.semalytix.com/): Overview of Semalytix and Project Atlas.
- [Platform](https://www.semalytix.com/platform): How Project Atlas turns patient voice into structured signal.
- [Disease areas](https://www.semalytix.com/disease-areas): Therapy areas Semalytix has worked across.
- [Company](https://www.semalytix.com/company): About Semalytix, the team, and the mission.
- [Contact](https://www.semalytix.com/contact): Reach the Semalytix team.

## Products

- [PV Sentinel](https://www.semalytix.com/safety-sentinel): Pharmacovigilance signal detection from the patient voice.
- [Outbreak Sentinel](https://www.semalytix.com/outbreak-sentinel): Public-health outbreak signals from open conversation.
- [Adherence Sentinel](https://www.semalytix.com/adherence-sentinel): Early-warning system for adherence and persistence.

## Knowledge Base (blog)

Field notes, methods and case studies on Patient Experience Data. Index: [https://www.semalytix.com/knowledge-base](https://www.semalytix.com/knowledge-base). RSS: [https://www.semalytix.com/knowledge-base/feed.xml](https://www.semalytix.com/knowledge-base/feed.xml).

- [LLM4KMU: Semalytix joins NRW-funded project bringing open-source LLMs to SMEs](https://www.semalytix.com/knowledge-base/llm4kmu-semalytix-joins-nrw-ai-project-for-smes) — Semalytix is a partner in LLM4KMU, a Bielefeld University–coordinated project funded by North Rhine-Westphalia to help small and medium-sized enterprises adopt open-source Large Language Models. (Semalytix Team, 2026-07-09, ~1 min)
- [POMELO: the standard patient-voice AI has been waiting for](https://www.semalytix.com/knowledge-base/pomelo-the-standard-patient-voice-ai-was-waiting-for) — Without scientifically sound frameworks, patient-voice AI is just louder guesswork. Why Semalytix co-developed the POMELO protocol, and how Atlas is engineered to it by design. (Philipp Cimiano, 2026-05-12, ~2 min)
- [What Patient Experience Data really is, and why it isn't a survey](https://www.semalytix.com/knowledge-base/what-ped-really-is) — PED is the unprompted, longitudinal voice of the patient. We unpack the difference between solicited PRO data and the messy, vital signal patients leave behind every day. (Janik Jaskolski, 2026-04-22, ~3 min)
- [Listening at scale: from a million forum posts to a single decision](https://www.semalytix.com/knowledge-base/listening-at-scale) — How agentic pipelines collapse months of manual coding into hours, without flattening nuance. A walkthrough of our ingestion → signal → action loop. (Janik Jaskolski, 2026-03-11, ~3 min)
- [The four adherence drivers we keep finding in 2026](https://www.semalytix.com/knowledge-base/adherence-drivers-2026) — Cost, side effects, perceived efficacy and access. The proportions shift by therapy area, but the four are remarkably stable. Here's what the cohort data shows. (Philipp Cimiano, 2026-02-04, ~3 min)
- [Why claims data is always two months late](https://www.semalytix.com/knowledge-base/claims-vs-conversation) — The structural reasons claims will never catch a discontinuation in time, and what teams should be measuring upstream of the refill gap. (Philipp Cimiano, 2026-01-19, ~2 min)
- [Regulators are warming to PED. What evidence do they actually want?](https://www.semalytix.com/knowledge-base/regulators-on-ped) — FDA, EMA and PMDA are all signalling. But the bar for inclusion in a label is specific: provenance, traceability, and a defensible cohort definition. (Janik Jaskolski, 2025-11-30, ~2 min)
- [Tolerability shifts show up in language before they show up in AEs](https://www.semalytix.com/knowledge-base/tolerability-shifts-early) — A pattern we've seen across three specialty therapies: patient vocabulary around side effects re-organises 4–8 weeks before the AE curve bends. (Philipp Cimiano, 2025-09-22, ~2 min)
- [Why we ship fleets of agents, not bigger models](https://www.semalytix.com/knowledge-base/fleets-not-models) — The case for narrow, accountable agents that hand off to each other, and why that architecture suits regulated pharma workflows better than a single oracle. (Janik Jaskolski, 2025-08-05, ~2 min)
- [Routing PSP outreach by reason, not by risk score](https://www.semalytix.com/knowledge-base/psp-routing-by-reason) — A field study with a mid-size brand: when patient-services teams act on the cause of disengagement instead of a generic risk band, lift compounds. (Philipp Cimiano, 2025-06-17, ~2 min)
- [Anonymisation without flattening: keeping the patient in the data](https://www.semalytix.com/knowledge-base/anonymisation-without-flattening) — How we strip PII while preserving the lived-experience texture clinicians and brand teams actually need to act. Pipelines, not promises. (Philipp Cimiano, 2025-05-02, ~3 min)
- [The silent $637B: putting a number on what we never hear](https://www.semalytix.com/knowledge-base/the-silent-637b) — An updated estimate of the global non-adherence loss, broken down by therapy area, with the assumptions made transparent so teams can challenge them. (Janik Jaskolski, 2025-03-14, ~3 min)