pai
A reusable multi-agent memory architecture in TypeScript for a local-first Personal AI platform.
TypeScript / Multi-Agent Systems / Event Sourcing / MCP
Suraj Lab
Suraj · Software Engineer at Amazon
I work on production backend systems at Amazon. Separately, I build side projects like pai, coding-agent, and a local-first knowledge base for memory, orchestration, and agent workflows.
All systems are made of tradeoffs. Document them.
Role
Software Engineer
at Amazon
Focus
Backend systems, distributed services, and agent infrastructure
Experience
10 years building backend systems
Base
Seattle, Washington, United States
01 / Work
Current work.
The projects and systems threads I want people to understand first.personal-knowledge-base
A local-first knowledge base with markdown notes in `kb/`, a shared core library, a Next.js UI, a `kb` CLI, MCP tools, optional S3 sync, and local semantic search.
Next.js / TypeScript / MCP / CLI / AWS S3 / Transformers.js
Memory Policy
How memory persists, flows, and stays inspectable across multiple agents.
TypeScript / Memory / Evaluation / Retention
coding-agent
A Claude Code plugin for building software end to end with five agents, deterministic gates, runtime evaluation, and a dedicated diagnosis mode.
Claude Code / Markdown / Bash / Playwright / Context7
02 / Writing
Writing, in reading order.
Short essays grouped by problem: continuity, memory, and reliable agent workflows.013 min read
Software Is Shifting From Tools to Systems
The real change is not that models can answer better. It is that useful software now needs loops for observation, correction, and carried state.
Systems With Continuity023 min read
The Demo Works. The Fifth Run Doesn’t.
First-run output is the easiest benchmark. Repeated work is where weak continuity shows up.
Systems With Continuity033 min read
Most AI Agents Don’t Improve Over Time
Repeated work should compound. Most agents still recompute.
Systems With Continuity043 min read
Context Is Not Memory
Memory that affects behavior needs status, provenance, correction, contradiction handling, and revocation.
Systems With Continuity03 / Experience
Experience.
A compact view of the professional background behind the project work.Amazon · Software Engineer
Backend engineering across retail, AWS, and robotics teams, with a focus on services that stay reliable under real operational load.
The University of Texas at Arlington · MS in Computer Science
Graduate study in computer science, with the foundations that later shaped my work in distributed systems, cloud infrastructure, and software architecture.
Software engineer at Amazon. Outside work I build tools around memory, orchestration, and reliable software.
- Production systems need explicit contracts, observability, and operational rigor.
- Side projects should solve real systems problems, not just stage demos.
04 / Contact
Good conversations start with a concrete systems problem.
Useful topics: backend architecture, local-first AI tools, agent infrastructure, memory systems, and orchestration.Agent-ready content
Structured profile data for tools, crawlers, and search without crowding the human-facing page.