Backend Engineer · California, USA
Amish Kushwaha
Applying distributed systems discipline to production AI infrastructure.
I build resilient backend systems, moving seamlessly between enterprise scale and greenfield startups. From optimizing Go control planes serving 25k+ virtual services at VMware / Broadcom to building AI gateways and agentic runtimes. My approach centers on strict observability, decoupled design, and fault-tolerant architecture.
/ What I Build
Engineering determinism into non-deterministic systems.
Bringing the rigorous architecture of enterprise control planes to the chaotic data demands of LLM orchestration and agentic state management.
Distributed Systems
High-performance control planes and distributed state management. Engineered to maintain strict consistency, low latency, and operational predictability across massive, multi-tenant network clusters.
AI Infrastructure
Provider-agnostic routing engines, decoupled RAG architectures, and high-throughput LLM gateways. Built to abstract model providers, handle failures gracefully, and optimize cost and performance at the network layer.
Agentic Runtimes
ReAct-style tool execution loops, MCP layers, sandboxes, and automation schedulers - the plumbing that lets agents do real work reliably.
/ Experience
4 years of systems.
BlueFunda
Software Engineer · Los Angeles, CA, USA · Feb 2025 – Present
Architected the foundational infrastructure for a high-throughput AI gateway and execution platform. Designed and implemented custom LLM routing engines, decoupled RAG pipeline architectures, and multi-turn agentic runtimes. Built entirely provider-agnostic and horizontally scalable from day one, enabling seamless, zero-overhead integration of external vector databases, model providers, and Model Context Protocol (MCP) servers.
VMware (Broadcom)
Member of Technical Staff · Bangalore, India · Jan 2022 – Nov 2024
Engineered core components within the distributed control plane of NSX Advanced Load Balancer. Spearheaded a complete redesign of the State Cache Manager concurrency model, built multi-region configuration versioning for Global Server Load Balancing (GSLB), and optimized the platform's distributed Resource Manager to surface highly responsive, real-time telemetry across the cluster architecture.
/ Build Log
Current systems in the workshop.
Things I build to understand systems better or because the right tool didn't exist yet.
Go daemon that scans and tracks ATS directly, filters jobs against a target profile using an LLM, and delivers Slack alerts the moment matching roles go live.
Per-ATS goroutine scheduling, SQLite dedup, exponential backoff, TUI audit browser. Deployed on a personal VPS.
/ Writing
Writing from the edge of the work.
Notes on backend systems, LLM infrastructure, and the non-obvious details behind reliable engineering.