AI Systems Engineer

Prateek Kumar Goel

Building AI systems that scale — from training clusters to production inference. Currently shipping open-source tools for agent reliability and RAG evaluation.

Open to meaningful challenges

PyTorch ·FSDP ·LLM Inference ·RAG ·Agentic AI ·MCP ·vLLM ·CUDA

Book a call View my work →

Currently building

Sits in front of AI agents to prevent hallucinations and maintain codebase consistency using a JIT Context Engine.

TypeScriptJIT Context EngineDAG SwarmLLM Agents

github.com/pkgprateek/setu-opencode

Runs RAG configurations head-to-head on the same queries using an LLM-as-judge panel for faithfulness and relevance.

PythonFastAPIChromaDBLLM-as-judge

github.com/pkgprateek/rag-arena-2026

More work →

Recent writing

The Intent Gap: Why the Best Coders Are Getting Outpaced by Average Engineers in 2026

Apr 17, 2026 6 mins read

The version of 2026 where the best programmer wins - NEVER HAPPENED"

software-engineeringai-agentscareersystem-designjobs-2026agentic-ai

Beyond the Wrapper: Why System Design & Architecture is the Highest Leverage Skill in the AI Era

Mar 25, 2026 8 min read

The most dangerous engineer in your organization right now is not the one who doesn't understand AI. It's the one who understands AI and nothing else.

system-designsoftware-architectureai-engineeringllmdistributed-systemsbackend-engineeringengineering-strategytech-leadershipragllm-orchestrationai-infrastructurecareer

Context Rot: Why Your AI Agent Gets Dumber the Longer It Runs

Mar 18, 2026 6 min read

After 60% context window usage, your AI agent enters the dumb zone - still confident, quietly wrong. Here's what that costs in production and how to fix it.

ai-agentsai-automationllm-hallucinationcontext-rotllmopsai-agent-memory

All writing →