powered by Cognera T2 ^(-d) & OpenAI GPT-5 ^Pro

Advanced AI - made with privacy in mind.
Made in Germany.Cognera has been updated to 2.1. Check the Changelogs for additional information.

• Chat now

We don’t monetize you – we empower you.

Most AI companies collect everything about you. We don’t.
Bytepersona is built on the principle of Datenminimierung – data minimization by design – going beyond GDPR requirements.

---

Our core values:

1. Minimal data collection

2. User experience first

3. Radical transparency

We log the bare minimum – no chat history exploitation, no hidden tracking, no profiling.Fast responses, clean interface, open model choice – without sacrificing privacy or raising costs.No dark patterns, no upselling tricks. Our policies are written for humans, not for lawyers.

Hosted in Germany.

Lake of endless possibilities

Our entire infrastructure is hosted in the Lake Constance region, one of Germany’s top-ranked areas for quality of life.Key benefits:
• Ultra-low latency: Direct, high-speed connectivity within central Europe ensures minimal delays for all users.
• Strict GDPR compliance: All data is processed and stored exclusively in Germany, fully protected by EU and German data protection laws.
• Full data sovereignty: You retain complete control over your data at all times—no third-party transfers outside the EU.*^{*Outside models from OpenAI, Anthropic, and Google, as well as those from other companies, have their own GDPR and data processing policies. We are purely handling the requests and anonymizing your user data.}
Note: Models from OpenAI and Google will soon be hosted on data servers owned by their respective companies within the EU.

• Details

Incredibly smart.

GPT-5^Thinking integration - by default

GPT-5 Thinking is OpenAI’s cutting-edge reasoning engine, purpose-built to interrogate large-scale datasets, detect intricate patterns, and extract actionable insights from complex documents.

• Details

Made in Germany.

Cognera 2

Bytepersona’s privacy-first mental-wellbeing engine has evolved from v1 to v2.
The new release builds on the solid foundation of v1, expanding the training corpus from 500 GB to >1 TB of rigorously curated psychological and behavioural-health data.Just like before, all data is housed exclusively on our own on-premises servers in Konstanz (Baden-Württemberg) — still no cloud hand-offs, no third-party access. With v2 you now get:
• Lower latency through optimized inference pipelines.
• Stronger GDPR alignment with improved compliance logging.
• Deeper insights from expanded behavioural-health coverage.The result: faster, more accurate, and still uncompromisingly private support for your mental well-being.

• Details

No where to be seen.

With privacy in mind

“Privacy is not for sale, and human rights should not be compromised out of fear or greed.” - Pavel DurovAt Bytepersona we live by that principle: your data is yours - period.What that means in practice
• No selling, sharing, or re-using We never monetise your chats or file attachments, nor do we feed them back into our training pipelines.
• No retention without consent By default nothing is stored long-term. You can opt-in (or back out) at any time for
• keeping chat transcripts,
• keeping file attachments,
• or deleting everything immediately after processing.
• GDPR-compliant erasure on demandIf you want your data gone, we invoke a BSI-2011-VS secure-wipe procedure, the method endorsed by Germany’s Federal Office for Information Security. Once finished, you receive a formal deletion certificate documenting time, method, and cryptographic checksums.In short, you remain in full control from upload to final deletion.

Free by default.

Essential - for well-being

Bytepersona is, and will remain fully functional at no cost. Every feature that powers our platform is included in the free tier.To preserve this model for all private users while deterring automated scraping and other abuse, we enforce fair-use rate limits. These ceilings are generous for individual use and are never invoked during normal interaction.If your project or organisation requires heavier throughput (e.g., large-scale analytics, continuous API access, or multi-seat deployments), expanded quotas and priority bandwidth will soon be available through paid plans.

Easy to use.

LibreChat - the easy to use AI interface

Bytepersona is powered by LibreChat—an open-source, ChatGPT-compatible front-end renowned for its clean UI and plug-and-play extensibility. Its modular architecture lets us add bespoke features without sacrificing usability, so if you’ve ever used ChatGPT you’ll feel right at home: same shortcuts, same conversation flow, zero learning curve.

Blueprint of our SoC Inferra 16

Introducing our new server-optimized chip designed specifically for efficient local hosting of large language models (LLMs). Featuring 32+ high-performance ARM/RISC-V cores, integrated AI accelerators, and 16GB high-bandwidth RAM, it delivers exceptional AI inference power while maintaining industry-leading energy efficiency. The blueprint has been sent to a leading semiconductor manufacturer in Dresden, Germany, for fabrication. Perfect for running LLMs without external knowledgebases, this chip sets a new standard for compact, sustainable AI in the data center.EDIT (9. SEP 2025): MANIFACTURING OF PROTOTYPE HAS STARTED!

Bytepersona is a registered trademark.
Located at 78467 Konstanz, Baden-Württemberg, Germany.

Where a high quality of life isn't just a claim - it's the proven result of a world-class economy, cutting-edge technology, and a deep-rooted commitment to our people.
This is Baden-Württemberg.

Sale Volumes

YTD Revenue: € 904.860
YTD Expenses: € 639.713
Operating Income (proxy): € 265.147
Operating Margin (YTD): 29,30 %
Average Monthly Revenue (9M): € 100.540
Average Monthly Expenses (9M): € 71.079
Run-Rate (ARR, 9-month average): € 1,206 M
Run-Rate (ARR, trailing 3 months Jul–Sep): € 1,556 M (Ø € 129.647/mo)
Run-Rate (ARR, trailing 6 months Apr–Sep): € 1,379 M (Ø € 114.910/mo)
CMGR (Jan→Sep): ~11,90 % (compound monthly growth)
Revenue multiple (Sep vs Jan): 2,46× (€ 146.950 vs € 59.780)
Quarterly revenue: Q1 € 215.400, Q2 € 300.520, Q3 € 388.940
QoQ growth: Q2 vs Q1 +39,52 %, Q3 vs Q2 +29,42 %
Expense ratio (YTD): 70,70 % (Expenses / Revenue)
Revenue volatility (σ monthly): ~€ 25.709 (CV ≈ 25,6 %)
Operating income trend: € 12.454 (Jan) → € 51.570 (Sep) (≈4,14×)

• Return

Legal Notice

Legal information according to § 5 DDG, § 18 MStV.Operator / Responsible for Content:
Bytepersona GmbH i. G.
Line-Eid-Straße 6
78467 Konstanz
GermanyManaging Director: Daniel MladekBranch in Dresden, Germany: {SOON}
Branch in Lisbon (Lisboa), Portugal: {SOON}Contact:
E-Mail: [email protected]

Liability for Contents
As a service provider, we are responsible for our own content on these pages in accordance with § 7 para. 1 DDG and general laws. According to §§ 8–10 DDG, however, we are not obligated to monitor transmitted or stored external information or to investigate circumstances that indicate illegal activity.
Obligations to remove or block the use of information under general laws remain unaffected. Liability in this regard, however, is only possible from the time of knowledge of a specific infringement. Upon notification of such violations, we will remove this content immediately.Liability for Links
Our offer contains links to external websites of third parties on whose contents we have no influence. Therefore, we cannot assume any liability for these external contents. The respective provider or operator of the pages is always responsible for the content of the linked pages. The linked pages were checked for possible legal violations at the time of linking. Illegal contents were not recognizable at the time of linking. A permanent control of the linked pages, however, is not reasonable without concrete evidence of an infringement. Upon notification of violations, we will remove such links immediately.Copyright
The contents and works created by the site operators on these pages are subject to German copyright law. Duplication, processing, distribution, or any form of commercialization of such material beyond the scope of copyright law requires the prior written consent of the respective author or creator. Downloads and copies of these pages are only permitted for private, non-commercial use.
Insofar as the content on this site was not created by the operator, the copyrights of third parties are respected. In particular, third-party content is marked as such. Should you nevertheless become aware of a copyright infringement, we ask you to inform us accordingly. Upon notification of violations, we will remove such content immediately.

• Return

About Bytepersona

Bytepersona began as a passion project in Vienna (2024) when Daniel (already captivated by the rise of ChatGPT 3.5-turbo) decided to channel his enthusiasm for applied AI into a real-world tool.
In November 2024 Jessica joined as lead backend engineer and co-owner, accelerating the platform’s technical growth.Today both founders call Konstanz on Lake Constance (≈ 88 000 inhabitants) home, a region renowned for its quality of life. Early-2025 marked a milestone: the Austrian sole proprietorship Bytepersona e.U. was migrated into a newly registered German company to support our growing international user base.Key dates
• Jan 2024 — Bytepersona prototype launched in Vienna
• Nov 2024 — Jessica joins, becomes lead of Operations Team
• Jan 2025 — Move to Konstanz & on-prem infrastructure setup
• Apr 2025 — German entity officially registered
• Jun 2025 — Jonas joins, handling legal matters & AI developmentWe remain a tight, founder-led team, focused on privacy-centric AI tools and guided by curiosity rather than venture capital.Our team:
Daniel - Founder of Bytepersona, AI Engineer
Jessica - Operations Lead, Backend and AI Engineer
Jonas - AI Engineer, legal team

Our team spans diverse roots. Daniel started in Lower Austria (2nd shape) before settling in Baden-Württemberg, Germany (1st shape). Jessica brings New Zealand heritage (3rd shape) to the mix, while Jonas also calls Baden-Württemberg home. A small world, yet an ambitious team.

• Return

Credits & Acknowledgements

Thanks to: zu0t, Boris, Cracker, wendego.Special thanks to: our IT equipment, stackoverflow, our family and friends. ❤️Bytepersona staff:
Jessica, Jonas & DanielThey've supported us for:

• Return

Privacy Policy

Datenschutzerklärung (DE)Die deutsche Version ist rechtsverbindlich.1. GeltungsbereichDiese Erklärung gilt für bytepersona.eu und chat.bytepersona.eu sowie alle damit verbundenen Online-Dienste („Plattform“). Sie erläutert, welche personenbezogenen Daten wir verarbeiten, zu welchen Zwecken, auf welcher Rechtsgrundlage und wie lange.2. Begriffsbestimmungen (Art. 4 DSGVO)„personenbezogene Daten“, „Verarbeitung“, „Verantwortlicher“, „Auftragsverarbeiter“ etc. richten sich nach Art. 4 DSGVO.3. VerantwortlicherBytepersona GmbH i. G. (notariell eingetragen in Freiburg i. Breisgau)
Line-Eid-Straße 6, 78467 Konstanz, Deutschland
E-Mail (Datenschutz): [email protected]Datenschutzbeauftragter: Derzeit nicht benannt, da keine gesetzliche Pflicht besteht. Wir prüfen die Voraussetzungen nach Art. 37 DSGVO fortlaufend.4. Rechtsgrundlagen der Verarbeitung
• Art. 6 Abs. 1 a DSGVO (Einwilligung), z. B. bei optionalen Diensten/Analytics/externe KI-Modelle.
• Art. 6 Abs. 1 b DSGVO (Vertrag), z. B. Registrierung/Account, Chat-Bereitstellung.
• Art. 6 Abs. 1 c DSGVO (Rechtspflicht), z. B. handels-/steuerrechtliche Aufbewahrung.
• Art. 6 Abs. 1 f DSGVO (berechtigte Interessen), z. B. IT-Sicherheit, DDoS-Abwehr, Stabilität.
• Endgerätezugriffe: § 25 TDDDG (ehem. TTDSG): Speicherung/Auslesen auf deinem Gerät (Cookies/Local Storage) nur, wenn unbedingt erforderlich oder mit Einwilligung. (Seit 14.05.2024 TTDSG → TDDDG umbenannt.)5. Datenkategorien & Zwecke
• Registrierung/Account: E-Mail, Passwort (gehasht), Verifizierungsstatus → Konto-Bereitstellung (Art. 6 Abs. 1 b).
• Nutzungs-/Sitzungsdaten: Session-ID, Sprache, Lizenzschlüssel, Nutzungszähler → Authentifizierung, Quotas, Komfort, Missbrauchsprävention (Art. 6 Abs. 1 b/f; § 25 Abs. 2 Nr. 2 TDDDG).
• Server-Logs: IP-Adresse, Zeitstempel, Request-URI, Referrer, User-Agent → Sicherheit/Fehleranalyse (Art. 6 Abs. 1 f).
• Chat- & Upload-Inhalte: Prompts, Antworten, hochgeladene Dateien → Kernfunktion (Art. 6 Abs. 1 b). Keine Nutzung deiner Inhalte zum KI-Training; Trainingsdaten stammen ausschließlich aus Mitarbeiter-Kurationsprozessen (kein Nutzer-Opt-in/-out nötig).
• Support-Daten: Kontaktmails inkl. Inhalt → Bearbeitung deiner Anfrage (Art. 6 Abs. 1 b/f).5a. Endgerätezugriffe (TDDDG § 25)Wir setzen technisch notwendige Cookies/Token ein, um den ausdrücklich gewünschten Dienst bereitzustellen (Login-Session, Bot/WAF-Schutz, Consent-Speicher). Nicht notwendige Endgerätezugriffe (z. B. optionale Analytics-Skripte) laden wir erst nach Einwilligung über das Banner.5b. Cookies/Token – Überblick

Hinweis: Cloudflare-Tokens sind funktional erforderlich für DDoS-/Bot-Abwehr und WAF-Challenges; sie tracken nicht sitzungsübergreifend.6. Empfänger / Auftragsverarbeiter / Drittlandübermittlungen
• Hosting: Hetzner Online GmbH (Falkenstein/DE) als Auftragsverarbeiter (Art. 28).
• E-Mail-Versand: Proton AG (CH) – Schweiz besitzt Angemessenheitsbeschluss (Art. 45 DSGVO).
• CDN/WAF/Edge-Security: Cloudflare, Inc. (USA). Übermittlungen stützen wir auf den EU-US Data Privacy Framework-Angemessenheitsbeschluss (Art. 45); Cloudflare ist im DPF gelistet. 
• Externe KI-Modelle: OpenAI OpCo L.L.C. (USA), Anthropic PBC (USA), Google LLC (USA) – nur bei aktiver Modellwahl; Rechtsgrundlage für die Übermittlung: EU-Standardvertragsklauseln (SCC, Beschluss (EU) 2021/914) + zusätzliche TOMs (Pseudonymisierung, TLS, rollenbasierte Zugriffe). 6a. Cloudflare-Metriken
1. Sicherheits- & Betriebsmetriken (Edge): Verarbeitung technischer Verkehrsdaten (IP, UA, Ländercode, Pfade, Statuscodes, Zeit) zur Angriffsabwehr, Caching/Performance, Verfügbarkeit. Rechtsgrundlagen: Art. 6 Abs. 1 f DSGVO (legitimes Interesse an Sicherheit/Performance) + § 25 II Nr. 2 TDDDG für die erforderlichen Endgerätezugriffe.
2. Cloudflare Web Analytics (optional, cookielos, kein Fingerprinting): Laden wir nur nach Einwilligung; erfasst z. B. Seitenaufrufe, Pfade, Ladezeiten. Laut Cloudflare keine Cookies/LocalStorage und kein Fingerprinting. Rechtsgrundlage: Art. 6 Abs. 1 a DSGVO. Widerruf jederzeit im Banner.7. Speicherdauer & Löschung
• Server-Logs: i. d. R. 14 Tage, danach Löschung/Anonymisierung (längere Aufbewahrung nur bei Sicherheitsfällen).
• Chats & Uploads: In deinem Konto bis zur Löschung; automatische Löschung inaktiver Verläufe nach 90 Tagen ohne Zugriff.
• Account-Daten: Löschung auf Antrag; Backup-Retention max. 30 Tage.
• Support-Mails: i. d. R. 6 Monate nach Abschluss des Vorgangs.
• Gesetzliche Belege: 6/10 Jahre nach HGB/AO (sofern angefallen).8. Technische & organisatorische Maßnahmen (Art. 32 DSGVO)Transportverschlüsselung (TLS 1.3, HSTS), Härtung/Patching, rollenbasierte Zugriffe (Least-Privilege), Cloudflare WAF/Bot-Management, Zero-Trust-Segmente, regelmäßige Pen-Tests, Incident-Response-Prozess.9. Betroffenenrechte (Art. 12 ff. DSGVO)Auskunft, Berichtigung, Löschung, Einschränkung, Datenübertragbarkeit, Widerspruch (Art. 21), Widerruf von Einwilligungen. Antwortfrist i. d. R. 1 Monat.
Beschwerde: Der Landesbeauftragte für den Datenschutz und die Informationsfreiheit Baden-Württemberg (LfDI BW), Lautenschlagerstraße 20, 70173 Stuttgart, Postfach 10 29 32, 70025 Stuttgart, [email protected], Tel. +49 711 615541-0.10. Automatisierte Entscheidungen / ProfilingKeine Entscheidungen i. S. v. Art. 22 DSGVO mit Rechtswirkung oder ähnlicher erheblicher Beeinträchtigung; Sicherheits-/Abuse-Filter sind funktional automatisiert, aber ohne Rechtswirkung.11. Pflicht zur BereitstellungE-Mail & Passwort sind vertraglich erforderlich (ohne Registrierung keine Nutzung). Externe Modelle/Analytics sind optional.12. Minderjährige / AltersgrenzeBei Diensten der Informationsgesellschaft erfordert eine Einwilligung grundsätzlich 16 Jahre; Deutschland hat die Grenze nicht abgesenkt (Art. 8 DSGVO).13. ÄnderungenMaterielle Änderungen geben wir bekannt; aktuelle Fassung steht auf bytepersona.eu.⸻Terms of Use (EN)The German version prevails in case of discrepancies.1. AcceptanceBy creating an account or using chat.bytepersona.eu, you accept these Terms and the Privacy Policy.2. Service & ScopeBytepersona provides a LibreChat web interface to interact with multiple AI models (internal and third-party). Access requires registration and email verification.3. EligibilityYou must be 16+ and legally capable of entering into contracts. Accounts registered by bots or automated methods are prohibited.4. Account & SecurityKeep credentials confidential; you are responsible for all activities under your account. Notify us immediately of any unauthorised use ([email protected]).5. Plans & AccessThe platform is primarily B2B. Advanced and Supersonic plans are available only to companies, agencies, or public authorities and are activated via license key or bespoke agreement. No classic subscriptions; Bytepersona does not store payment card data on its systems.6. Acceptable UseDo not upload/produce illegal or infringing content; do not interfere with security or traffic-management features (including Cloudflare WAF/Bot Management), circumvent challenges, scrape, overload, or reverse-engineer the service. Temporary access restrictions (challenge pages) are part of our protective measures.7. Intellectual PropertyThe platform software, UI/UX, and internal models are owned by Bytepersona or its licensors. We grant a limited, non-exclusive, revocable licence to access the service for its intended purpose. No reverse engineering or derivative works.8. User Content & OutputsYou retain ownership of your prompts/uploads and outputs. You grant Bytepersona a non-exclusive, worldwide, royalty-free licence to process/store/transpose your content solely to provide the service. We do not use user content to train models.9. External ModelsWhen you select a third-party model (OpenAI, Anthropic, Google), your content is transmitted to that provider under their terms. Bytepersona is not liable for third-party processing once data leave our infrastructure.10. AI TransparencyWe may label AI-generated content where misidentification as “real” is likely; users must not remove such labelling. Chat interactions clearly indicate you are interacting with AI.11. Disclaimer of WarrantiesService is provided “as is” without warranties of any kind, express or implied, including availability, accuracy, or fitness for a particular purpose.12. Limitation of Liability (DE carve-outs)We are liable without limitation for intent and gross negligence, injury to life/body/health, product liability, and where we explicitly assumed a guarantee or acted fraudulently.
For simple negligence in breaching material contractual duties (Kardinalpflichten), liability is limited to the foreseeable, typical damage. Otherwise, liability is excluded. Mandatory consumer rights (including Sections 327 et seq. BGB for digital products) remain unaffected.13. IndemnificationYou shall indemnify Bytepersona from third-party claims arising from your unlawful use or infringement of these Terms, including reasonable legal costs.14. TerminationWe may suspend/terminate with prior notice for violations or misuse; you may delete your account at any time via the UI or by email.15. ModificationsWe may modify or discontinue the service. Material changes to these Terms will be notified in advance; if you object within 6 weeks, either party may terminate.16. Governing Law & JurisdictionGerman law applies (conflict-of-law rules excluded).
Consumers: mandatory consumer protections of your residence remain; venue is your local courts.
Businesses: exclusive venue is Stuttgart, Germany.17. SeverabilityIf any provision is invalid, the remainder stays in force.18. Contact[email protected]

• Return

Training Data of
Cognera T1^(-d)

700 GB — transcripts of therapy sessions and online interactions
This extended dataset expands the textual base with more diversity of sources and deeper annotation.
├── 500 GB: transcripts of over 200,000 hours of psychotherapy sessions (multiple therapeutic schools, disorders, demographics). Each session tagged with multi-label annotation: speaker roles, therapy style, ICD-10/DSM-5 categories, emotional trajectory.
└── 200 GB: large-scale collections of online support forums, chat transcripts, blogs and long-form narratives, filtered for psychological relevance. Annotated for emotional polarity, coping strategies, relapse indicators, and longitudinal behaviour tracking.350 GB — Audio and video analysis of psychological interactions
The multimodal base has been upgraded with higher-resolution capture and richer annotation.
├── 250 GB: audio recordings (>40,000 hours) covering natural conversations, structured interviews, voice diaries. Annotated for prosody, micro-pauses, disfluency markers, and biometric vocal stress cues.
└── 100 GB: video datasets (>750,000 clips/images) of faces, microexpressions, gestures, and body posture. Includes 3D facial landmarking, skeletal tracking, and temporal emotion dynamics.150 GB — Scientific literature and specialised articles
Now broader and more interlinked with structured references.
└── Over 3.5 million documents: peer-reviewed psychological and psychiatric articles, clinical case reports, therapeutic manuals, meta-analyses, and longitudinal cohort studies. Enriched with knowledge graph extraction, connecting theories, interventions, and outcomes.

• Return

ToS & Privacy Policy for "Cognera" Discord Bot Instance

Synthara Bot – Terms of Service (ToS)
Last updated: 30 June 20251. PURPOSE
Synthara Bot ("Bot") provides AI‑powered slash‑commands (including /gpt and /synthara) on Discord. These Terms govern your access and use of the Bot’s services ("Service"). By invoking or interacting with the Bot you agree to be bound by these Terms.2. DEFINITIONS
• Developers – Daniel M. & Jessica P.
• User – Any Discord account interacting with the Bot.
• Premium / Pro – Paid or otherwise granted tier that unlocks additional AI models.3. ELIGIBILITY & ACCESS
You must (a) be 16 years or older (or have guardian consent), (b) have a valid Discord account, and (c) not be on the Bot’s blacklist.4. ACCEPTABLE USE
You agree NOT to:1) Violate Discord’s Terms of Service or Community Guidelines.
2) Request or generate content that is illegal, hateful, or violates third‑party rights.
3) Reverse engineer, scrape or overload the Service.
4) Circumvent paywalls or token limits.
Developers may rate‑limit, throttle or terminate access at any time to maintain Service integrity.5. PREMIUM & MONETIZATION
Premium ("Pro") features are optional and delivered after successful payment processed via Discord’s monetization platform.
• Prices and benefits are displayed before purchase.
• Subscriptions auto‑renew monthly until cancelled.
• EU 14‑day digital‑content withdrawal is waived upon first execution of Premium features.6. DISCLAIMERS
The Service is provided “AS IS”. Outputs are generated by third‑party AI models (OpenAI & Together AI) and may be inaccurate or biased. Developers make NO WARRANTIES regarding correctness, reliability or fitness for a particular purpose.7. LIMITATION OF LIABILITY
To the maximum extent permitted by law, Developers shall not be liable for indirect, incidental, special or consequential damages arising out of or related to your use of the Service. Direct liability is capped at the amount you paid in the preceding 12 months.8. TERMINATION
Developers may suspend or terminate your access at any time for breach of these Terms or abuse of the Service. You may stop using the Bot at any time; data retention is governed by the Privacy Policy.9. GOVERNING LAW
These Terms are governed by German law. Exclusive venue is Konstanz, Germany, unless mandatory consumer‑jurisdiction rules apply.10. CHANGES TO TERMS
We may modify these Terms with at least 14 days’ notice posted via the Bot’s /about command. Continued use after the effective date constitutes acceptance.11. CONTACT
Email: [email protected]---Synthara Bot – Privacy Policy
Last updated: 30 June 20251. DATA CONTROLLER
Daniel M. & Jessica P., Konstanz – Germany (“we”, “us”). Contact: [email protected]2. WHAT DATA WE COLLECT

3. LEGAL BASES (GDPR)
• Art. 6(1)(b) – Contractual necessity: deliver requested AI responses.
• Art. 6(1)(f) – Legitimate interest: prevent abuse, maintain Service integrity, usage metrics.
• Art. 6(1)(c) – Legal obligations (accounting & tax for paid features).4. THIRD‑PARTY RECIPIENTS
Provider | Data shared | Safeguards
OpenAI, LLC (USA) | Prompt & context for GPT models | SCCs, DPA; retained ≤30 days
Together AI, Inc. (USA) | Prompt for perplexity‑ai/r1‑1776 | SCCs, DPA
Discord Inc. | Messages via Discord infrastructure | Discord ToS & Privacy Policy5. RETENTION
• Token logs: 12 months, then aggregated/anonymised.
• Whitelist / blacklist / Pro flags: until deletion request or account removal.
• Prompts & outputs: not stored beyond processing (max 24h temp cache for debugging).6. YOUR RIGHTS
GDPR rights: access, rectification, deletion, restriction, portability, objection. Contact us (§1). Complaints → German DPA / Baden-Württemberg (https://www.baden-wuerttemberg.datenschutz.de/).7. SECURITY MEASURES
• TLS enforced; minimal collection; short retention.
• Production keys accessible only to Developers (2FA + hardware token).8. CHILDREN
Service not directed to children under 16. If we process such data without consent it will be deleted.9. CHANGES TO THIS POLICY
Updates announced via /about. Material changes may require renewed consent.10. CONTACT
Email: [email protected]

Current job listings

AI Engineer (m/f/d)

Job Description: AI Engineer (m/f/d) – Focus on Psychological AI (Compliant with German & EU Law)Company: Bytepersona
Location: Konstanz, Germany / Remote within Germany or EU
Employment: Full-time (part-time negotiable)
Start: flexible

About Us

Bytepersona is a dynamic, fast-growing startup based in Germany, dedicated to creating AI-driven solutions for psychological research and real-world applications. As a young company at the intersection of technology and mental health, we combine the flexibility and creativity of a startup with the responsibility of strict compliance with German and European law—including the General Data Protection Regulation (GDPR/DSGVO), the German Federal Data Protection Act (BDSG), and anti-discrimination laws (AGG). Our small, interdisciplinary team is passionate about shaping the future of psychological AI responsibly and making a positive impact on mental health through transparent, ethical technology. If you want to build something new, take ownership, and grow with us, you’ll fit right in.

Your Responsibilites

• AI System Development & Optimization
Design, train, and evaluate machine learning/deep learning models for psychological data (e.g., language, text, sensor streams), in full compliance with GDPR and local data protection requirements.• Modeling Psychological Constructs
Implement validated psychological theories (e.g., emotion, personality, stress models) into robust, transparent, and auditable AI algorithms.• Data Integration & Preprocessing
Collect, process, and analyze multimodal data (questionnaires, voice, wearables) ensuring anonymization, consent management, and highest standards for data security and privacy.• Interdisciplinary Collaboration
Cooperate closely with psychologists, data scientists, legal/privacy officers, and software engineers to ensure both scientific quality and legal compliance.• Model Validation, Monitoring & Fairness
Conduct feasibility studies, validate models in practical settings, and proactively monitor for bias, fairness, and clinical relevance as required by German and EU regulatory frameworks.• Documentation & Transparency
Comprehensive documentation of AI development, model decisions, and data processing procedures—facilitating regulatory audits and scientific publication.

Your Profile

• Master’s/PhD in Computer Science, AI, Data Science, Mathematics, Computational Psychology, or a related discipline.• Solid programming skills (Python, R, Julia, etc.) and practical experience with leading ML frameworks (PyTorch, TensorFlow, scikit-learn).• Experience in developing, evaluating, and deploying AI models (preferably in language, text, or psychological domains) in regulated environments.• Familiarity with GDPR/DSGVO, BDSG, and European AI ethics guidelines (see EU AI Act), or willingness to acquire this knowledge.• Awareness of risks regarding bias, explainability, and the psychological impact of AI; proactive, responsible approach to risk assessment.• Strong analytical skills, structured working style, and team communication—fluent in English (German skills are a plus or must be willing to learn).

What We Offer

• A legally compliant, ethically responsible work environment with a focus on mental health and technological progress.• Flexible working models (remote/hybrid possible within the EU, core team based in Germany).• Ongoing legal and technical training (including GDPR/AI Act compliance workshops).• Modern technical infrastructure and high-performance computing resources.• Open, diverse team with a clear anti-discrimination policy according to the German General Equal Treatment Act (AGG).• Transparent processes, flat hierarchies, and open feedback culture.

Legal & Ethical Notice

We value diversity and provide equal opportunities in accordance with German and European law (AGG, GDPR, EU AI Act). All applications—regardless of gender, origin, disability, age, identity, or background—are encouraged. Sensitive data is processed exclusively according to current data protection laws.

Salary

€4,642.86 ^gross / mo.

The minimum annual gross salary for this position is €65,000, paid out in 14 installments per year (12 monthly salaries plus additional vacation and Christmas bonuses, each equivalent to one monthly salary).
This means your gross salary per payout is €4,642.86.
The total annual amount remains €65,000, in accordance with the Austrian model of salary distribution. Actual compensation depends on qualifications and experience; we are open to negotiating a higher package for outstanding candidates.Note:
Inspired by the Austrian salary model, we have voluntarily adopted this structure for our German startup. Although German law does not require the division of annual salary into 14 payments, we offer this benefit to provide additional financial flexibility and to acknowledge candidates familiar with the Austrian system. All standard taxes and social security contributions apply.

• Apply Now

Application Form for:
AI Engineer (m/f/d)

Your data is processed exclusively for the purpose of the application procedure and stored for 6 months.
After this period, all application data and related communications will be deleted, unless you have given explicit consent for longer storage. You have the right to request information about your data, correction, deletion, restriction of processing, objection to processing, and data portability as provided by law.If you have given consent for extended storage, you may withdraw it at any time with effect for the future. You also have the right to lodge a complaint with a data protection supervisory authority. For further information, please see our Privacy Policy.

Changelogs

Cognera 2

Scroll down to reveal previous versions

2.1 – 28/09/2025Added
• Multimodal v2.1 pipeline: Expanded to support audio embeddings (speech-to-reasoning).
• Fine-grained access controls in the Plugin SDK (per-plugin sandboxing, resource quotas).
• Lightweight client SDKs for mobile (Swift + Kotlin, alpha).
• Incremental training hooks for online adaptation in closed-test deployments.Improved
• Multimodal performance: Image + text fusion latency reduced by ~35%.
• Symbolic Reasoning module: Optimized for distributed inference; throughput gain ~25% on multi-node clusters.
• Context handling: Stable up to 512K tokens, adaptive truncation beyond that point.
• API layer: Lower round-trip time on gRPC, ~20% faster under high concurrency.Fixed
• Occasional deadlocks in adaptive batching under multi-tenant workloads.
• Tokenizer regressions with non-Latin scripts (Arabic, CJK).
• Inconsistent logging metadata in OpenTelemetry traces.Known Issues
• Audio embeddings are limited to English (multilingual support planned for 2.2).
• Incremental training only available in staging; production rollout pending compliance review.
• GPU memory fragmentation may occur in multimodal + symbolic combined workloads; monitoring required.> Responsible: JP (pri.), JB (sec.), DM (ter.)

2 - 23/09/2025Added
• New Neural Core v2.0 with extended context window and hybrid reasoning (transformer + symbolic inference).
• Adaptive batching engine for dynamic load balancing in multi-user environments.
• REST / gRPC / WebSocket APIs shipped by default.
• Plugin SDK (Python, Rust, Node.js) for custom integrations.
• Knowledge graph engine for structured reasoning and toolchain expansion.
• Initial multimodal pipeline (text + image embeddings, beta).Improved
• Inference speed up to +40% faster vs. Cognera 1 due to optimized GPU kernels (CUDA / ROCm).
• Memory efficiency and reduced latency in high-concurrency workloads.
• Security hardening: sandboxed plugin execution, stronger encryption defaults (TLS 1.3).
• Observability stack: full OpenTelemetry support (metrics, traces, logs).Fixed
• Reduced token overflow errors in long-context scenarios.
• Stability improvements for distributed inference across GPU clusters.
• Authentication bugs in SAML / OAuth2 enterprise integrations.Known Issues
• Multimodal features are experimental - currently under optimization; expected improvements in v2.1.
• Symbolic reasoning module not yet fully optimized for large-scale distributed deployments.
• Throughput drops beyond 256K token context length.> Responsible: JP (pri.), JB (sec.), DM (ter.)

• Return

Contact us