Three Patent-Pending Products · One Adaptive Agentic Framework

Adaptive Agentic AI
for High-Stakes Domains

Activense AI builds three patent-pending Adaptive Agentic AI products on a single deterministic-core, policy-orchestrated, human-governed framework — for metabolic health, cognitive rehabilitation, and conscious wellbeing. Safety is structural, not aspirational.

DIPP145329 Recognised DPDP-Aligned Architecture India Residency Human-in-the-Loop by Design Bootstrapped · Self-Funded
3Patent-pending products
4Patent applications
4Trademarks filed
11Microservices in prod
9,000+Clinical interactions (ground truth)
0Autonomous dosing · ever

Built by a clinical co-inventor who codes the safety in

Dt. Shilpa Bhoskar — Founder & Co-Inventor

Dt. Shilpa Bhoskar

Founder · Co-Inventor on All 4 Patent Applications · Director, Activense Holistic Solutions Pvt Ltd

Founder and co-inventor across all three AI products. Co-invented and authored the safety architecture (ADR-005 Structural Dosing Prohibition, ADR-004 Confidence-Gated Escalation, ADR-009 Deterministic-Core Agentic Framework) that powers the Activense AI portfolio.

Clinical credentials provide the ground-truth labelling discipline behind every product: Gold Medal — MSc Dietetics & Food Service Management · MBA · Certified Nutritionist & Sports Nutritionist · NLP Practitioner · Visiting Faculty, Bharati Vidyapeeth, Pune.

"Safety in clinical AI cannot live in a prompt. It must live in the schema, the database trigger, and the policy engine. That is how we built it."
shilpabhoskar@activense.com · +91-9923218618 · LinkedIn available on request

Three products. Four patent applications. One framework.

Each product is independently patentable and serves a distinct high-stakes domain. They share an underlying technical platform — the Adaptive Agentic Framework — which is what makes safety, audit, and policy guarantees consistent across the portfolio.

Live · 2 patents pending
Glycosutra
Adaptive Agentic AI for Pre-Diabetes Reversal

India's first patent-pending agentic AI platform for diabetes prevention. Multimodal meal capture, real-time CGM correlation, clinician-in-the-loop policy engine. Live in production with a dietitian-supervised cohort. Patent-pending Low-GI atta flour (GI 48-52) and phenotype-aware substitution engine across 621 Indian foods in 7 languages.

27 patent claims Multimodal capture CGM correlation 621 Indian foods CGM optional
Visit glycosutra.ai →
Live · 1 patent pending
Cognitrion
Adaptive Agentic AI for Aphasia & Cognitive Rehabilitation

India's first AI-powered aphasia and neuropsychological assessment platform. Built on clinical neuropsychology expertise from the University of Bristol and University of Stirling. Structured assessment, progress tracking, and personalised rehabilitation pathways for stroke and TBI patients. Safety architecture inherited from the shared framework.

Aphasia Stroke rehab Neuropsychology TBI Patent pending
Visit cognitrion.ai →
Sprint 0 · 1 patent pending
Sankalpa-i
9-Agent Consciousness Companion AI

A mobile-first AI consciousness companion rooted in Vedic wisdom. 9-agent Navagraha-mapped architecture, 4 sacred time windows, 7 Solfeggio frequencies mapped to the chakra system. The highest-stakes member of the portfolio for safety — protected by the Guardian Gate and Adaptive Escalation Protocol. 108 founding seekers. PCT planned.

9-Navagraha agents React Native Vedic AI 108 seekers PCT planned
Visit sankalpai.in →

The Adaptive Agentic Framework

A patent-claimed architectural framework that powers all three products. Safety, audit, and policy are designed as structural guarantees — not prompt-level instructions. The same Architecture Decision Records govern every product in the portfolio. Implementation details are patent-claimed and available under NDA.

ADR-009 · LOCKED

Deterministic core · LLMs as components

The policy engine is the sole decision authority. Large language models are bounded components for capture and narration — never controllers. No tool-calling agents with side effects. No autonomous workflow agents. No self-directed loops.

ADR-003

Tamper-evident audit architecture

Every clinical decision recorded in a tamper-evident audit trail. Append-only at the data layer, exportable for regulatory review. Specific implementation is patent-claimed; details under NDA.

ADR-004 · PATENT-CLAIMED

Confidence-gated human escalation

Multi-tier human escalation with confidence-gated routing. Low-confidence outputs never act autonomously — they request human review. The specific tier architecture is patent-claimed; mechanism details available under NDA.

ADR-005 · ABSOLUTE

Structural safety prohibitions

No endpoint, schema, model field, or UI surface in the entire codebase can express medication or insulin dosing. Enforced at the architecture level, not at the prompt level. Carries across every healthcare product in the portfolio.

ADR-010 / 011 · PATENT-CLAIMED

Provider-neutral LLM router

Provider-neutral routing across Gemini · Claude · GPT-4o · Llama, selected per call by task and safety profile. A patent-claimed multi-model safety architecture governs high-stakes decisions; mechanism details under NDA.

Human Governance

Human-in-the-loop by design

Dietitians and clinicians are the sole approval authority for patient-facing changes. Agents are advisory only. LLMs are components, never controllers. Self-directed or goal-seeking agents are forbidden by architectural decree.

A coherent IP portfolio

Indian provisional patent applications across all three products, one underlying technical framework, and formal Startup India recognition.

4
Patents pending
2× Glycosutra · 1× Cognitrion · 1× Sankalpa-i
4
Trademarks filed
Activense · Glycosutra · Cognitrion · Sankalpa
DIPP
145329
Startup India recognition · Sept 2023
27+
Patent claims (Glycosutra alone)
Provisional GANMS-2026-001 · Filed 31 Mar 2026

How the AI portfolio fits within the parent group

We surface this structure proactively because credibility in a regulated AI product portfolio depends on clean disclosure, not concealment.

Current entity structure

  • Parent entity — Activense Holistic Solutions Pvt Ltd · Pune, India · CIN U74999PN2022PTC216787 · Pvt Ltd incorporated December 2022 (business operating since 2013 under earlier brands). Owns the patents, signs vendor contracts, engages the team, and operates production today.
  • AI products R&D arm — Activense AI (this portfolio, activense.ai). Three patent-pending Adaptive Agentic AI products on a shared technical framework. DIPP145329-recognised by Startup India.
  • Clinical practice — Activense (activense.com). The 10+ year clinical-domain practice that funds the AI R&D, provides ground-truth labelling for the healthtech products, and acts as the human-in-the-loop oversight for Glycosutra and Cognitrion in production.

We welcome questions from investors, partners, and programme reviewers about how this maps onto your evaluation criteria.

Get in touch

For partnerships, clinical collaborations, investor enquiries, or programme reviewers.

Lead Contact

Dt. Shilpa Bhoskar shilpabhoskar@activense.com

+91-9923218618

Registered Office

Activense Holistic Solutions Pvt Ltd

CIN U74999PN2022PTC216787
Warje, Pune, Maharashtra 411052
DIPP145329