Aliri+ is the conversational AI layer on top of Aliri NLP and your BI feeds. Ask questions in plain English — across clinical findings and operational metrics — and get cited, grounded answers in seconds. No SQL, no dashboards, no waiting. Deployable on-prem when PHI demands it.
What was the average TAT for studies flagged with a critical finding last quarter, by shift?
Median TAT on critical-finding studies was 38 min overall — 27 min on day, 51 min on night.
Median TAT (min) by shift
You
How has the rate of incidental thyroid nodules on CT chest changed over the last 12 months, and which radiologists most often recommend follow-up?
Aliri+
Across 14,283 CT chest reports in the last 12 months, incidental thyroid nodules were mentioned in 9.4% — up from 7.8% in the prior year. Follow-up recommended in 62% of those.
Top three recommenders this quarter: Dr. Patel (74%), Dr. Nguyen (69%), Dr. Okafor (67%).
Generated Chart
live previewIncidental thyroid nodule rate
CT chest · monthly
% of reports mentioning a thyroid nodule
Now
9.4%
12mo ago
7.8%
Δ
+1.6pp
From the Product
Two real conversations with Aliri+ — one zoomed into a single patient's clinical trajectory, one across a department's operational dataset. Same product, same plain-English Q&A, no SQL written by hand.
Querying 21,148 rows across 1 table — AI will run SQL on demand.
Based on the records for patient bc60c5a4dea8, 72d002123d99, there appears to be a significant progression and fluctuation in findings, particularly regarding a right-sided pneumothorax (collapsed lung) and associated pleural effusions, tracked over a period in April 2016.
Key observations from the reports include:
In summary, the patient experienced a dynamic clinical course with a worsening and then fluctuating pneumothorax requiring multiple chest tube interventions, alongside other abdominal and systemic findings.
Querying 11,004 rows across 2 tables — AI will run SQL on demand.
Here are the average Turnaround Times (TAT) by procedure type for the top 20 procedures:
| Procedure Description | Total Exams | Average Total Time (HH:MM:SS) | Min Total Time (HH:MM:SS) | Max Total Time (HH:MM:SS) |
|---|---|---|---|---|
| CHEST - 2 VIEWS | 1,020 | 8:35:13 | 00:00:00 | 4:00:54:00 |
| CHEST - 1 VIEW | 426 | 7:35:27 | 00:00:00 | 22:15:00 |
| MAM BILATERAL SCREENING DIGITAL MAMMOGRAM WITH TOMO | 419 | 29:49:53 | 07:07:09 | 2:06:59:42 |
| CT HEAD WITHOUT CONTRAST | 311 | 9:30:44 | 00:00:00 | 3:19:47:00 |
| US ABDOMEN AND PELVIS LIMITED | 292 | 12:15:21 | 00:00:00 | 7:10:09:00 |
| CT Ang Chest w and/or w/o Contrast | 274 | 00:00:00 | 00:00:00 | 00:00:00 |
| …14 more rows | ||||
On open, Aliri+ summarises the dataset — what tables, how many rows, what time window. The user sees the scope before asking anything.
"Show me the average TAT by procedure type." Aliri+ generates the SQL on demand and returns a structured answer — table, totals, ranges.
Aliri+ proposes follow-ups grounded in what it knows about the data — TAT distribution, top institutions, longest workflow steps, trend over time.
How It Works
Aliri+ doesn't read raw reports. It reasons over the structured clinical and operational layers you've already built — so answers stay faithful to your data.
Entities, findings, anatomy, modifiers — extracted from free text and coded to RadLex and SNOMED CT. The result is a queryable clinical layer.
RVU, turnaround time, shift schedules, modality, radiologist assignments, study volumes — connected alongside the clinical data.
A large language model is grounded in this unified structured layer — so answers cite real concepts, real rows, real time windows. No hallucinated medicine.
Deployment
Connect Aliri+ to a commercial frontier model for aggregate, de-identified workloads — or use the Aliri-provided on-prem LLM appliance running on Apple Silicon when PHI safety is non-negotiable.
Claude · ChatGPT · Gemini
Connect Aliri+ to the frontier model of your choice. To keep PHI safe, only aggregate, de-identified data is sent to the commercial LLM — never patient identifiers. Ideal for operational BI, cohort metrics, and trend analytics.
Apple Silicon · Aliri-provided
For PHI-sensitive deployments, Aliri ships an integrated on-prem LLM running on Apple Silicon — you don't choose, manage, or fine-tune the model. We provide it, optimized for radiology workloads, behind your firewall. Patient identifiers never leave your environment.
Privacy by design
Aliri+ inherits the same privacy posture as the rest of the Aliri stack. In Commercial mode, only aggregate, de-identified data is sent to the cloud LLM — PHI is never transmitted, which limits analytics to non-patient-level questions. In Local mode, every token of PHI stays on the Apple Silicon appliance behind your firewall — full patient-level analytics, no data ever leaves your network.
Two Domains, One Conversation
Aliri+ reasons across both layers — clinical findings from Aliri NLP and operational metrics from your BI feeds — in the same question.
Clinical Intelligence
Built on Aliri NLP's ontology-coded entities — anatomy, findings, modifiers, negation, certainty. Ask about prevalence, trends, cohorts, follow-up gaps, or critical-finding patterns.
Example question
“Show me all patients with a 5-10mm pulmonary nodule on a CT in the last 6 months whose follow-up CT was never completed.”
Operational Intelligence
Connects to your BI feeds — radiologist productivity, turnaround time, shift schedules, modality mix, volumes. Ask workforce, throughput, and quality questions in plain English.
Example question
“Which radiologists had the longest TAT on weekend night shifts last quarter, and how did their RVU/hour compare?”
Capabilities
Every capability is grounded in your existing structured data. No new ontologies, no new pipelines — just a smarter way to ask.
Describe a cohort in plain English — Aliri+ resolves it to RadLex/SNOMED concepts and runs the query.
Ask how a finding rate, modality mix, or operational metric has shifted across time windows.
Join clinical entities with operational data in one question — e.g. "TAT vs. critical-finding rate by shift."
Every answer carries citations back to the structured rows, reports, and ontology codes it used.
Aliri+ proposes visualizations for the answer — bar, line, distribution — ready to pin to your Visualize dashboards.
Follow-up questions retain context: refine the cohort, change the time window, drill into a subgroup.
Reason over RVU, TAT, shift assignments, and radiologist mix — operational questions without writing SQL.
Push the resulting cohort, table, or chart back into your BI tools, EHR worklists, or research workflows.
Get Started
See how Aliri+ turns your radiology and operational data into an AI analyst your team can talk to.
