Technology Stack
Cutting-edge technologies powering the future of AI-driven care.
LUKi is built on a modular architecture with selective open-source components and enterprise-grade services. The platform prioritizes privacy, modularity, and deployment flexibility across on-premises and cloud environments.
Core Languages & Libraries
Primary Development
Python- The backbone language for AI/ML development and data processing
Natural Language Processing
spaCy- Industrial-strength NLP parsing (entities, segmentation).Hugging Face Transformers- State-of-the-art transformer models and utilities.NLTK- used only for legacy/tokenization tasks.
Machine Learning & AI
PyTorch 2.x– Primary framework for inference and fine-tuning.scikit-learn– Classic ML utilities (clustering, anomaly detection, trend analysis).LangChain– Limited use for tool decorators only; custom agent orchestration with FastAPI.Sentence-Transformers– Embedding generation for ELR® chunks.
Data Processing & Storage
pandas- High-performance data manipulation and analysisChromaDB / FAISS- Local vector stores for long-term memory retrieval.SQLite / Postgres– Lightweight structured data storage for metadata/logs.
AI Platforms & Models
Large Language Models (LLMs)
Open‑source large language models—accessed via providers such as Together AI—serve as the primary conversational and generative backbone, with the stack designed to be easily upgraded as models improve.
Custom Fine‑tuned Adapters (LoRA/QLoRA) – Domain adaptation to ELR® and caregiving context.
Cloud/Compute Infrastructure
Together AI hosted API for primary model inference. Roadmap includes switching to local hosting once MVP beta testing is complete.
Containerized deployment via Docker, with support for orchestration on modern Kubernetes platforms (e.g. EKS/GKE) and other managed cloud environments used during the beta phase.
Future support for self-hosted GPU infrastructure
Blockchain & Decentralization
Oracle Networks
Chainlink - Secure connection between on-chain smart contracts and real-world data
Custom Oracle Solutions - Healthcare-specific data feeds and verification
REMEGrid Infrastructure
Convex-based Lattice - High-throughput, low-cost transaction environment
Smart Contracts – CAP→REME conversion, data-access permissions, RUBI distribution.
Database & Storage
Graph Databases
Neo4j / NetworkX – Social graph modeling, interest matching, community linkage.
Distributed Storage
IPFS - Decentralized file storage for healthcare records
Encrypted Data Lakes - Compliant archival of anonymized aggregates.
Security & Privacy
Encryption & Privacy
AES‑256 encryption is applied to sensitive ELR data at rest, with TLS used for all service‑to‑service and external communication as the default protection layer.
Federated learning (future roadmap) with secure aggregation & differential privacy.
Differential Privacy & Secure Aggregation – Protect individuals in aggregated model updates.
Zero‑Knowledge Proofs / Homomorphic Encryption – Roadmap items for advanced verifiability.
Compliance & Governance
Architecture designed in line with HIPAA healthcare data protection principles.
Systems engineered with GDPR requirements in mind, with formal compliance work treated as an ongoing, continuously improving process.
Audit Trail Technology - Transparent, immutable record keeping.
High Performance • Security First • Globally Scalable
Last updated