Suraj · Backend Engineer
A systems-minded backend engineer building towards an agent-native future, not just using AI tools but architecting the memory, orchestration, and tooling layer beneath them.
After 8 years building production backend systems at Amazon, I’m moving deeper into the stack: asking what software architecture looks like when agents are first-class citizens, not bolted-on features.
Current Work
The current thread is a reusable multi-agent memory architecture in TypeScript: event-sourced, local-first, and built for agents that need durable context instead of prompt-sized hacks.
A reusable multi-agent memory architecture in TypeScript for a local-first Personal AI platform.
Current project focused on memory, correction, and context accumulation.
Designed around memory as a state machine, append-only logs, and separation of read and write concerns so agents can preserve context without collapsing into hidden state.
TypeScript · Multi-Agent Systems · Event Sourcing · MCP
Approach
I care about durable systems, explicit state, and infrastructure that remains inspectable once agents are in the loop.
Foundation
Backend / Distributed Systems / ScaleEight years at Amazon building systems that had to work at scale. Distributed systems, operational rigor, and the discipline to distinguish prototypes from production.
Focus
Agents / Memory / ArchitectureBuilding reusable multi-agent memory architecture in TypeScript, using event-sourced ideas, append-only logs, read and write separation, search, and MCP tooling.
Method
Strategy / Systems Thinking / First PrinciplesThe same rigor applied to code shows up everywhere else: decomposing systems, challenging defaults, and building independent models instead of borrowing inherited assumptions.
Writing
The writing archive is split into two tracks: how software becomes a system with continuity, and what it takes for agent-driven software to actually ship.
Archive
Each piece lives in its own static page. New essays only require adding another entry to writings.txt; the archive and routes update from that file at build time.
2
Reliable Agent Systems
Systems With Continuity
4 essaysEssays on memory, governed state, adaptation, and why useful software should compound over time.
01
4 min readThe most important change in software is not just better models. It is the shift from one-shot tools toward systems that observe, remember, adapt, and operate over time.
02
4 min readMany AI products are powerful in the moment and forgetful over time. That is why they often feel better in demos than in repeated use.
03
4 min readTools, prompts, and workflow steps can make agents more capable. They do not automatically make them adaptive.
04
4 min readWithout durable, governed memory, AI systems can be capable without really becoming more useful over time.
Reliable Agent Systems
5 essaysEssays on coding agents, verification, orchestration, and the structure required to ship working software.
05
4 min readCode generation is local. Shipping software is end-to-end. A lot of coding agents still confuse the two.
06
3 min readThe more important a rule is, the less it should depend on prompt wording alone.
07
3 min readThe question is not whether your coding system has one agent or five. The question is whether each piece of structure isolates a real failure mode.
08
5 min readA case study in separating roles, adding deterministic gates, forcing runtime verification, and building just enough memory for the next session to start smarter.
09
4 min readThe most important lessons were not about prompts. They were about structure, verification, memory, and the limits of self-evaluation.
Experience
Eight years of backend engineering at Amazon, followed by a deliberate move deeper into the stack: asking what software architecture looks like when agents are first-class citizens, not bolted-on features.
Timeline
Worked across various teams at Amazon since 2016, spanning warehouse automation, fulfillment backend systems, AWS and CloudFront-related infrastructure, and Amazon Robotics automation. Built and operated distributed backend services with a focus on scale, reliability, and production rigor.
Graduate study in computer science, with the foundations that later shaped my work in distributed systems, cloud infrastructure, and software architecture.
Approach
I work in the layer between product apps and ML research: the infrastructure and architecture layer for agentic systems.
Contact
I’m especially interested in conversations around backend architecture, agent infrastructure, memory systems, and orchestration.
Direct channels
Best fit
Usually replies within a few business days.
Email directly