PyTorchGPUAWS g5 / p4

Technology

We build with GPU inference and training. The stack is meant to run in the cloud, in production.

GPU compute

What runs in each product today

  • PyTorch with GPU acceleration for predictive model training
  • Inference on AWS GPU instances (g5, p4)
  • Embeddings and vector search (JurixSolve RAG)
  • Time-series models on regasification telemetry (GasSolve)
  • Population risk scoring (HealthSolve)
  • Production inference optimization (roadmap)

Machine Learning

  • PyTorch
  • Python
  • Scikit-learn
  • MLOps pipelines

NLP / Legal (JurixSolve)

  • GPU semantic embeddings
  • Vector search (FAISS)
  • Procedural RAG
  • Stage classification

Infrastructure

  • AWS (EC2 GPU, Lambda, S3)
  • Amplify + CloudFront
  • Amazon SES
  • CI/CD

GPU Compute

  • PyTorch + GPU
  • AWS instances (g5/p4)
  • Batch and online inference
  • Model monitoring

Workloads by product

  • Clinical and family risk prediction, city/department breakdown, and EPS profiling (HealthSolve)
  • Real-time pressure, temperature and IoT monitoring with time-series models and alerts (GasSolve)
  • Procedural drafting assistance and similar Colombian case-law retrieval (JurixSolve)
  • ETL pipelines for RIPS, industrial data, and judicial publications