Clinical decision support built to transform African healthcare, starting in Tanzania: offline, bilingual, and grounded in national guidelines.
Clinical decision support built to transform African healthcare, starting in Tanzania: offline, bilingual, and grounded in national guidelines.
Tanzania has roughly 1 physician per 10,000 people. Most clinical encounters happen at dispensaries and health centres staffed by clinical officers and assistant medical officers working with limited diagnostic equipment, inconsistent drug supply, and unreliable connectivity. The national Standard Treatment Guidelines exist, but they live in thick printed manuals that are rarely at hand during a consultation.
Meanwhile, most digital health tools are built for contexts that don't match this reality: they assume stable internet, English-only input, server-side processing, and clinicians with time to learn new workflows. When these tools arrive in a rural Tanzanian dispensary, they either don't work at all or they add friction instead of removing it.
That gap between what exists and what's needed is where ClerQ operates.
ClerQ is a Consumer Clinical Decision Support System built to transform African healthcare, starting with frontline clinicians in Tanzania. It is grounded in national Standard Treatment Guidelines (STG) and runs entirely on the phones clinicians already carry. No cloud dependency, no internet requirement, no new hardware.
When a clinician encounters a patient, ClerQ provides contextual guidance at the point of care. Not a generic chatbot answer. Not a web search. A reasoned clinical suggestion grounded in the protocols that clinician is already trained to follow, adapted to the drugs and resources actually available at their facility, and delivered in the language they and their patient speak.
Lucy is the AI assistant at the core of ClerQ's clinical decision support. She is the interface a clinician interacts with when they need guidance, and she is designed to do one thing well: take a clinical question in English or Swahili and return a grounded, traceable answer fast enough to be useful during a consultation, not after it.
Lucy doesn't generate prose. She doesn't hallucinate dosages. Every answer she gives can be traced back to the specific STG entry, formulary reference, or clinical protocol it was derived from. When she doesn't have enough information to answer safely, she says so. That's the design: correct over impressive.
Lucy runs entirely on-device. No server calls, no data leaving the phone, no dependency on connectivity. The full clinical knowledge base is embedded locally.
Ask in Swahili, get an answer in Swahili. Ask in English, get an answer in English. Lucy handles both without translation artifacts, powered by Sema's structured bilingual pipeline.
Every clinical suggestion maps directly to Tanzania's Standard Treatment Guidelines. Lucy doesn't improvise treatment plans. She surfaces the guideline-recommended approach for the presenting condition.
Every response includes the source reference. A clinician can verify Lucy's suggestion against the original guideline text. Nothing is a black box.
Before clinical guidance ever reaches a clinician's phone, it must be mathematically and clinically verified against national protocols. Hakiki is ClerQ's declarative validation engine. It audits every symptom weight, treatment threshold, and drug contraindication against Tanzania's Standard Treatment Guidelines (STG) and NEMLIT, chapter by chapter, with line-by-line page citations.
When guidelines change, we update declarative data files, not brittle application code. Hakiki also exposes an open Model Context Protocol (MCP) server, allowing authorized external AI tools (Claude, Cursor, Antigravity) to query Tanzania's STG guidelines directly under strict human-in-the-loop review.
Zero clinical logic hardcoded in software. Rules are structured datasets mapped verbatim to official STG sections and page numbers for auditability.
Confidence scores are not black-box machine-learned probabilities. They represent explicit bar-crossing ratios against clinician thresholds. Red-flag findings saturate immediately.
Decouples guideline reasoning from user interfaces. External models query the STG graph via MCP, proposing structured edits for expert verification.
A clinician's workflow with ClerQ is designed to add zero friction to existing practice.
A clinical officer at a dispensary sees a patient with a presenting complaint. They open ClerQ on the phone they already carry.
They describe the clinical scenario to Lucy in English or Swahili. Natural language, no structured forms, no dropdown menus. Just the clinical question as they'd describe it to a colleague.
Lucy returns a guideline-grounded suggestion within seconds. The response includes the recommended management approach, relevant drug dosages from the STG, and a direct reference to the source guideline section so the clinician can verify independently.
The clinician uses the suggestion alongside their own clinical judgment. ClerQ doesn't replace the clinician. It gives them the right information at the right time so they can make better-informed decisions under real-world constraints.
ClerQ is engineered with deliberate purpose to transform African healthcare, starting in Tanzania with a focus on real frontline clinical reality. We begin by solving the hardest operational constraints first: offline execution, low-resource language processing, and strict guideline alignment, establishing a foundation of trust before expanding across the continent.
Most AI tools work well in English and poorly in everything else. ClerQ treats Swahili as a first-class language, starting with Tanzania's linguistic context rather than an afterthought translation. Our Sema pipeline ensures that Swahili input receives the same quality of reasoning as English, without relying on machine translation that loses clinical nuance.
We don't build for the infrastructure we wish existed. We build for the infrastructure that's actually there: intermittent power, shared mobile devices, 2G connectivity in the best case, and nothing at all in the common case. ClerQ works fully offline because that's not an edge case in rural clinics. It's the default.
ClerQ's clinical knowledge base starts directly from Tanzania's Standard Treatment Guidelines and NEMLIT, rather than international protocols that may recommend drugs or procedures unavailable in-country. The system knows what's available, what's recommended, and what's realistic at the facility level.
Seven principles guide every technical and product decision at ClerQ. They're not aspirational values on a wall. They're engineering constraints we hold ourselves to, and they operate together, continuously, like bodies in orbit around the same center.
Hover to explore
Each principle orbits the same center: they work together, continuously.
Hover or tap the orbiting bodies
