With persistence and perseverance, small things make it BIG. I'm a Software Development Engineer II
at Amazon with 12+ years of experience building scalable systems. I work on advertisement optimization
and rendering engines serving over 1.1 trillion impressions annually worldwide.
My work spans machine-learning pipelines, distributed systems, and high-scale infrastructure. Beyond
coding, I'm an IEEE-published researcher in ML-based network security, and a photography enthusiast
capturing moments as ISerialClicker.
Years in
12+years
From embedded firmware to trillion-impression ad rendering — distributed systems, ML pipelines, and the occasional security paper.
By the numbers
12+
Years engineering
From research code to trillion-scale production systems
1.1T+
Ads served / year
Worldwide ad impressions through systems I help own
5B+
Daily peak impressions
Sustained throughput during Prime Day & Turkey 5
$450M+
Business value enabled
Through the high-scale pricing engine
40+
GitHub repositories
Open-source projects across AI tooling, infrastructure, and tools
1
IEEE publication
ML-based threat detection in SDN
Currently focused on
Machine-learning pipeline optimization
High-scale distributed systems
AI-powered developer-productivity tools
Open-source contributions
Work Showcase
GenAI Configuration Assistant
Hackathon winner — built end to end
AI-powered tool reducing manual configuration time from 2 hours to 2 minutes.
Tackled the challenge of translating natural-language requirements into complex system configurations using prompt engineering and iterative refinement.
1,300+ hours saved per year across the team
60× faster than the previous manual flow
98% accuracy on the validation suite
AWS BedrockClaude 3.5LambdaAPI GatewayTypeScript
High-Scale Pricing Engine
Distributed calculation system processing millions of real-time pricing decisions.
Solved challenges around eventual consistency, idempotency, and traffic spikes during peak events like Prime Day. Built on DynamoDB + Lambda with SQS dead-letter handling.
$450M+ business value enabled across 10+ regions
JavaDynamoDBLambdaSQSCloudWatch
International Expansion Platform
Multi-region launch for a new international market.
Led the technical design for launching services in a new market. Navigated regional compliance, data residency requirements, and adapting systems to different business rules.
Successful market launch with zero critical incidents
JavaAWSMicroservicesDynamoDBAPI Gateway
ThoughtMCP
MCP server implementing human-like cognitive architecture for AI reasoning.
Mimics biological cognitive processes through dual-process thinking, working memory, emotional processing, and metacognitive monitoring — giving AI agents richer, more transparent reasoning chains.
Toolkit for AI agents to follow best practices through structured specifications.
Comprehensive toolkit for AI agents to follow best practices through structured specifications, intelligent context detection, and automated validation. Provides modular instruction systems and enterprise-ready architecture for reliable AI agent workflows.
Hierarchical task management accessible via MCP, REST, and web UI.
Hierarchical task-management system designed for both humans and AI agents. Supports PostgreSQL and filesystem backends, enabling seamless task coordination across AI-assisted workflows.
Cloud Computing Specialization. Focused on distributed systems and scalable architecture.
GPA 3.74 / 4.0
Cloud Computing Specialization
Coursework in distributed systems and scalable architecture
B.E., Electronics & Communication Engineering
Gujarat Technological University
Foundation in embedded systems and signal processing.
Foundation in embedded systems and signal processing
Publications
Machine-Learning Based Threat-Aware System in Software-Defined Networks
Keyur Golani
Developed ML models integrated with an SDN controller for intrusion detection. Trained on ~150GB of network-traffic datasets. Improved routing in SD-WAN networks, reducing packet loss by more than 50%.
Transform Kiro into an expert UI/UX developer with comprehensive guidance on accessibility, design systems, component libraries, and modern UI development practices.
AgentSpec is a comprehensive toolkit that enables AI agents to follow best practices through structured specifications, intelligent context detection, and automated validation. Transform your AI development workflow with modular instruction systems and enterprise-ready architecture.
🧠 AI cognitive architecture over MCP — human-like five-sector memory, parallel reasoning, metacognition (bias/emotion/confidence), a 3D memory graph, and a web UI. PostgreSQL + pgvector + Ollama embeddings. 20 MCP tools.
🤰 A beautiful, interactive prenatal learning web app — 35 educational topics with audio narration, trimester-aware recommendations, a reflective journal, and calming light/dark themes. Helps expectant parents bond with baby during pregnancy. React + Node + MongoDB.
📝 A Notion-style, block-based note editor that replaces your browser's new tab — 18+ block types, slash commands, AI assist (OpenAI/Anthropic/Gemini/Ollama), daily notes, backlinks, and 100% local storage. Manifest V3.
🧠 AI cognitive architecture over MCP — human-like five-sector memory, parallel reasoning, metacognition (bias/emotion/confidence), a 3D memory graph, and a web UI. PostgreSQL + pgvector + Ollama embeddings. 20 MCP tools.
♟️ A modern, self-hostable checklist app for creating, sharing, and tracking reusable checklists. Template blueprints with nested references, tracked instances with progress, workspaces, collaboration, and a full Theme Studio. Next.js 15 + PocketBase.
🎨 Self-hostable, YAML-driven portfolio site with pluggable visual variants — write one portfolio.yml, swap the entire look without touching content. 28 section kinds, 5 themes, GitHub integration, and a browser admin editor. Next.js + TypeScript.