Johnvessly Alti

AGENTIC SYSTEMS • MCP SERVERS • RAG PIPELINES

About

I'm an AI Engineer who builds at the intersection of large language models, agentic systems, and production infrastructure. I design RAG pipelines and semantic vector search systems, orchestrate multi-step AI workflows with LangGraph, and expose AI capabilities as MCP servers secured with OAuth 2.1 so agents can interact with production systems directly. I ship full-stack TypeScript products — from NestJS backends with pgvector-powered retrieval to Next.js frontends — that go beyond demos into reliable, scalable systems. I think in systems: how context flows, where models fail, and how to engineer around their limits with the right retrieval, memory, and tooling strategies.

Technologies

Projects

DevNest
PG RAG

Open Source Contributions

1 / 2
SigNoz/signoz

fix(alerts): set default time range to 5m on create alert page #11601

Open

Identified ClickHouse Out-of-Memory (OOM) errors caused by default 6-hour high-cardinality queries during new alert creation. Optimized performance by reducing the default time range to 5 minutes, mitigating backend memory spikes in production scale environments.

TypeScriptReactClickHouseAlerting Systems
View Pull Request

GitHub Contributions

Consistent shipping — one commit at a time.

 

Blogs