Reading
AuraDent is a real-time documentation platform for dental clinics. Using Deepgram speech recognition, AI intelligent extraction, and AWS asynchronous processing, it automatically converts doctors' chairside dictations into structured medical records, treatment charts, and post-treatment guidelines.
Section 01
AuraDent is a real-time documentation platform for dental clinics. Using Deepgram speech recognition, AI intelligent extraction, and AWS asynchronous processing, it automatically converts doctors' chairside dictations into structured medical records, treatment charts, and post-treatment guidelines.
Section 02
During dental treatment, doctors need to record medical records, update treatment charts, and write post-treatment guidelines while treating patients—these documentation tasks are both time-consuming and error-prone. The traditional approach is to fill in records from memory after treatment, which makes it hard to ensure the accuracy and completeness of information. AuraDent was created to address this industry pain point: letting doctors focus on treatment while AI handles documentation.
Section 03
AuraDent uses a TypeScript monorepo architecture, integrating real-time voice processing, AI intelligent extraction, and asynchronous post-processing. The entire system is divided into five core modules:
Section 04
The real-time gateway built with Fastify and WebSocket is the system's entry point. It receives front-end audio streams from browsers and forwards them to Deepgram for speech recognition. The gateway manages session lifecycles, distinguishes between partial and final transcriptions, and performs PII (Personally Identifiable Information) desensitization before sending content to AI.
Section 05
This is the system's "brain", built on the Vercel AI SDK. The agent receives desensitized transcribed text, extracts structured clinical findings through typed tool calls and Zod validation. For example, when a doctor says "The patient's lower right second molar needs root canal treatment", the agent identifies the tooth position (#31), diagnosis (needs root canal treatment), and updates the corresponding data structure.
Section 06
The clinical terminal interface built with React + Vite provides real-time feedback to doctors. The interface includes:
Section 07
Responsible for converting the raw structured data extracted by the agent into a record format suitable for persistence, including deduplication logic (merging multiple mentions of the same tooth) and source tracing (recording the voice segment corresponding to each finding).
Section 08
A post-processing module based on AWS Lambda. When a session ends, the gateway sends session data (desensitized transcriptions, structured findings, tracking records, performance metrics) to an SQS queue, triggering the worker to generate post-treatment PDF guidelines, simulate insurance pre-authorization, and write the complete record to PostgreSQL.
Keep going with more reads from the same topic.
Splinter is a minimalist, high-performance key-value (KV) and vector storage system enabling zero-latency inter-process communication via shared memory and atomic operations. With only 766 lines of core code, it supports millions of operations per second and 768-dimensional vector storage, offering a new architectural approach for local LLM inference and data-intensive applications.
Folkering OS is the world's first AI-native bare-metal operating system, entirely written in Rust no_std without relying on Linux, POSIX, or libc. It can generate commands from scratch, compile them into WASM, and run them in 10 seconds, achieving true self-evolution.
This project explores how to use Large Language Models (LLMs) to detect logical vulnerabilities in Ethereum smart contracts, especially business logic flaws that are difficult for traditional static analysis tools to capture, such as economic manipulation and execution order errors.
This article introduces an open-source project for building modern large language models from scratch. The project evolves from a GPT-2-style basic architecture to a Llama 2/3-style production-level implementation, covering complete tutorial implementations of key technologies such as RMSNorm, RoPE, SwiGLU, GQA, and MoE.