AVORIX launched in 2026. These are our current and active engagements — each scoped with a clear architecture, defined outcomes, and full engineering ownership.
Designing a production RAG system to ingest large document corpora across regulatory filings, internal policies, and market research — with a target of sub-500ms query latency and high answer accuracy on domain-specific eval sets.
Hybrid vector + BM25 retrieval, re-ranking with cross-encoders, streaming responses via SSE.
Architecting a high-frequency data visualization platform for live market data, designed to handle large WebSocket event volumes with a multi-tenant structure and microsecond-accurate time series.
Go backend with actor model concurrency, Redis Streams for event sourcing, TimescaleDB for OHLCV storage.
Scoping a multi-agent system to classify, extract, and route legal documents at scale — with the goal of significantly reducing manual review time while maintaining high classification accuracy.
LangGraph multi-agent orchestration, custom fine-tuned classifier, async Celery pipeline with DLQ.
Building a fully automated Kubernetes platform across multiple AWS regions with zero-downtime deployments, automated failover, and full observability for a microservices-based SaaS product.
GitOps with ArgoCD, Crossplane for cloud resources, Istio service mesh, Karpenter for node autoscaling.
Scoping a compliant data platform to aggregate patient records from multiple EMR systems into a unified API — FHIR-native, with role-based access, full audit logging, and end-to-end encryption.
FHIR R4 server, event-sourced audit trail, field-level encryption via AWS KMS, GraphQL federation.
Building an ensemble ML system to forecast customer behaviour in advance, integrated into CRM workflows to trigger automated retention and personalisation campaigns.
XGBoost + LSTM ensemble, feature store on Snowflake, Airflow DAG for weekly retraining, REST API for CRM hooks.