Problems we solve:
- Data is often disorganized and unstructured (e.g., hand-written clinical notes), which makes it difficult to categorize information and derive meaning
- Without advanced data analytics, faint signals are hidden, and opportunities for information discovery, pattern recognition, and decision support can easily be missed
What we do:
- Establish a concrete understanding of the existing data model and determine the appropriate ML techniques to evaluate the data
- Apply Deep Learning (DL) and NLP to efficiently and accurately detect patterns in unstructured data, in order to extract actionable intelligence
- Train, tune, test, and validate predictive models that can be applied to additional data sets
- Analyze data – and produce visualizations and reports – for multilevel business consumption, so that analysts and executives can use their organization’s data to make better decisions
Impacts:
- Significant increases in efficiency, speed, and accuracy of data gathering and analysis
- Ability to evaluate system performance and forecast trends to support better decisions
- Increased productivity, decreased overhead costs, and optimized resource allocation based on variable assumptions and scenarios
Our expertise:
- Descriptive and predictive analytics
- Business models and rules engines
- Data integration
- Data curation and refinement
- Data quality
- Metadata management
- Anomaly detection and remediation
Technologies we use:
- ML/NLP frameworks: TensorFlow, PyTorch, scikit-learn, spaCy, NLTK, ktrain, huggingface, Prophet
- Data management platforms: Databricks, Palantir, TAMR
- Programming languages: Python, SQL, PySpark
- Visualization tools: Tableau, PowerBI, plotly, Matplotlib
- Libraries: Jupyter, Pandas, NumPy, Flask, SciPy
What we build or enhance:
- Dashboards, visualizations, and reports
- Analysis software/microservices
- Predictive ML/NLP models
- Workflows and business process improvements that use ML/NLP
- Rules engines
- Decision support tools and applications
Our Approach
Most healthcare providers have no systematic way to generate actionable insights from the billions of unstructured medical notes contained in Electronic Health Records (EHRs). Amida extends data analytics platforms to include NLP analysis of free-text clinical notes. We provide data processing automation, NLP models, and no-code (or low-code) sandbox environments for non-technical users. Applied to vast data sets, advanced NLP techniques help identify signs of serious medical conditions early enough to support meaningful interventions and improve health outcomes.
