Sound Prediction Inc. is a health technology company headquartered in Columbus, OH, operating virtually across the continental United States. We build Glance, a health data platform used by providers, ACOs, and individuals to aggregate, normalize, and act on medical records from any source.
We work closely with our research partners at Mederrata Research to ground our products in rigorous methodology and published clinical evidence.
The problem we’re solving
Healthcare data is broken in a specific, fixable way: every system records the same patient using different coding standards, different identifiers, and different formats. An Epic EHR speaks FHIR. A claims file speaks X12. A pharmacy uses NDC codes. A wearable uses its own schema. None of them talk to each other.
The result: providers make decisions on incomplete information. ACOs can’t close care gaps they can’t see. Patients can’t answer basic questions about their own health without calling three offices.
We build the infrastructure layer that fixes this — normalizing everything to the OMOP Common Data Model and delivering the analytics on top.
What we build
Glance is our flagship product. For providers and ACOs, it handles CMS DPC integration, SMART on FHIR EHR connectivity, quality measure computation, HCC risk coding, care gap identification, and financial decision support. For individuals, it’s a unified personal health record with lab interpretation, medication tracking, cost transparency, and true data ownership.
We help practices onboard into the CMS Data at the Point of Care program and handle the technical integration.
Predicato is our open-source temporal knowledge graph framework — a Go library for building knowledge graphs that evolve over time, with a fully local ML stack and no external API dependencies.
Our track record
We started in 2018 as Mederrata Inc., with a focus on using informatics to reduce medical error. In 2019, we were selected as semifinalists in the CMS AI Health Innovations Challenge — a competitive federal program identifying AI solutions with the highest potential impact on healthcare outcomes. At the time, our core team held positions at the National Institutes of Health.
During the COVID-19 pandemic, we began productizing our research. That work became Glance — a platform that brings together our expertise in clinical data standards, Bayesian modeling, and healthcare interoperability.
Our research has been published in ICLR, AISTATS, MLHC, and PLOS, covering interpretable models for readmission risk, discharge placement, and treatment effect estimation in clinical populations. We build from evidence.
Our core values
We are anti-AI hype
We are opposed to the externalities of the AI industry and the cynical hype used by AI companies to promote their products. The pattern is familiar: companies race to inject large language models into every workflow, promise magical results, and quietly send your most sensitive data to third-party cloud services to make it work. In healthcare, this means patient records flowing to companies whose business models depend on aggregating data at scale.
We refuse to participate in this.
That doesn’t mean we reject useful technology. We use small, purpose-built models for specific tasks where they genuinely help: language processors for clinical text understanding, OCR for document digitization, and entity recognition for extracting structured data from unstructured sources. These are tools — not oracles. They run locally in our application via Predicato and go-candle, or through a hosted service we provide and secure. They never call OpenAI, Google, Anthropic, or any third-party AI provider.
The distinction matters: a language processor that identifies “metformin 500mg” in a clinical note is a useful tool. A general-purpose LLM generating a “care plan” from a patient’s chart is a liability dressed up as innovation.
Privacy and safety first
Your data never leaves your control. Glance runs on your infrastructure — on-prem or your cloud. Vocabulary lookups run against a local DuckDB parquet file. Reference ranges are embedded in the binary. Scoring algorithms are pure math. The patient app stores data on-device. Nothing in Glance requires sending patient data to any third party — including us.
You choose the level of AI involvement. Purpose-built models can run entirely locally on your infrastructure with zero network calls. Alternatively, you can use a hosted service that Sound Prediction provides and secures. Either way:
- You control the granularity and de-identification level of any data processed
- You have full transparency into exactly what information is transferred over the network
- You can inspect every request and response for the hosted service
- You can turn it off entirely and the platform continues to function — just without NLP features
We don’t collect your data. Sound Prediction’s business model does not depend on aggregating or monetizing health information. We sell software. If you stop using Glance, your data doesn’t exist anywhere we can reach.
Deterministic where it matters. ETL mappings are reproducible SQL. HCC code derivation follows CMS-specified logic. Quality measure computation shows its numerator, denominator, and the HEDIS specification it implements. Care gap identification is rule-based. Financial projections use declared formulas. None of these are “AI-powered” — they are computed, auditable, and explainable to a clinician, a regulator, and a patient.
Open methodology. The ETL pipeline is inspectable. The care gap logic is documented. The scoring algorithms are transparent. The language models are identified by name and version. You don’t have to take our word that the results are correct — your team can read the code.
What we do
Interpretable modeling. Healthcare decision-making is high-stakes — it requires models you can explain to a clinician, audit for a regulator, and trust for a patient. We build transparent, interpretable algorithms for quality monitoring, risk adjustment, care gap identification, and financial decision support. No black boxes.
Health data interoperability. The 21st Century Cures Act requires payers to open claims databases, providers to share EHR data, and patients to delegate access to applications of their choice. Glance implements these standards — FHIR R4, SMART on FHIR, CMS Blue Button 2.0, CMS DPC — and normalizes everything to the OMOP Common Data Model.
Clinical research. Through our partnership with Mederrata Research, we publish peer-reviewed work on interpretable models for readmission risk, discharge placement, and treatment effect estimation. Our research grounds our products in methodology, not marketing.
Technology
Glance is built on open standards:
- OMOP Common Data Model for clinical data standardization
- FHIR R4 / STU3 for EHR interoperability
Get in touch
Whether you’re a medical director evaluating value-based care tools, a practice administrator exploring CMS DPC, or an individual who wants to own their health record — we’d like to hear from you.
Email: info@mederrata.com
