Production stack
The 15 technologies across our production systems. Every tool earned its place through real tradeoffs. No resume-driven choices. Click any item for engineering notes from the systems we ship with it.
15
Technologies
15
With full engineering notes
37+
Production systems
75+ Rust crates handle ingest, signal computation, and exchange execution with sub-100ms tick-to-signal latency.
TypeScript strict mode across 18 Next.js sites. Zero any-casts in production API routes.
18 Next.js 15 App Router sites covering marketing, SaaS dashboard, operator tools, and public APIs.
React 19 powers all client-side interactivity across the Avo platform, from the trading dashboard to the operator hub.
Node.js 20 powers the ClickHouse query proxy, API middleware, and server-side Redis caching in the Next.js layer.
Python handles IBKR integration via ib_insync, ML model training with XGBoost and LightGBM, and Claude API agents for outreach.
723M+ rows across 43 tables. Migrated from QuestDB in April 2026. 5-10x better compression, sub-second range queries on minute bars.
PostgreSQL handles transactional state: 25 tables, ~600K trades, lead pipeline, engagement metrics, and campaign state machines.
Redis 7.2 handles regime caching (209K keys, 24h TTL), signal pub/sub, job queues, and API rate limiting. Sub-millisecond p99 reads.
Kafka evaluated for the signal transport layer. Redis Streams chosen for Avo's own pipeline. Avo builds Kafka integrations for clients who already run it.
Docker stacks run Jitsi (video calls), PeerTube (video hosting), and Mattermost (team comms) as self-hosted alternatives to paid SaaS.
Kubernetes evaluated for Argus service orchestration. systemd on bare-metal chosen for Avo's own stack. Avo builds K8s-native platforms for clients on AWS, GCP, or Azure.
Vercel deploys all 18 Next.js sites. Preview deployments on every PR, production promotion via GitHub Actions gate.
Claude Haiku and Sonnet run across 4 production agents: email personalization, reply classification, market context synthesis, and lead scoring.
OpenAI is in the Avo client toolkit for vector search, vision tasks, and ChatGPT-compatible APIs. Avo uses Anthropic internally but knows when OpenAI is the better client-side call.
Looking for what we built, not what we build with?
Browse 37+ case studies →