Vamil Gandhi — Software Engineer

About

Based in

New York City, NY

Currently

Amazon · Software Engineer

Education

Carnegie Mellon University
M.S. Information Systems

Focus

Backend · AI Agents · MCP

Open Source

Actively building

I'm a software engineer building distributed backend systems, AI-enabled analytics platforms, and agentic workflows. At Amazon I architect systems used by 200K+ sellers daily. Outside work I build open-source AI tooling — from MCP servers to multi-agent orchestration frameworks.

Work

2023–Now

Amazon

Current

Software Engineer — New York City, NY

Built AI-powered analytics systems powering 200K+ users across Seller and Vendor Central — natural-language data retrieval, LLM-powered dashboard creation, and MCP servers for agent-driven workflows. Built reusable AWS CDK constructs cutting infra setup from 3 weeks to 1 day.

ReactTypeScriptJavaLangGraphMCPAWS CDKLangFuseCopilotKit

Feb–Jul 2023

Walmart Global Tech

Backend Engineer — Reston, VA

Backend APIs for Scan & Go checkout — $1.2M+ in daily revenue, 300K+ daily users. Led Ibotta integration and App2App auth features.

JavaAPIsBackend

Jun–Aug 2021

Walmart Global Tech

Software Engineer Intern — Reston, VA

Intelligent Self-Checkout monitoring via Dynatrace. Android Kotlin age-check feature covering 9M+ requests/month.

KotlinAndroidDynatrace

2018–19

Infosys

Systems Engineer — Pune, India

APIs and UI components with ReactJS and Python, improving response times by 20%. SAP SmartForms and ABAP Web DynPro automation.

ReactJSPythonSAP ABAP

Open Source

01Creator

AgentDispatch

A provider-neutral agent task dispatch system — abstracts over AWS, GCP, Azure, Kubernetes, and local runtimes behind a stable adapter contract. Ships with an MCP server, TypeScript SDK, CLI, and SQLite store so AI agents can dispatch, monitor, and cancel durable tasks across any cloud without vendor lock-in.

core mcp-server sdk-js cli store-sqlite adapter-aws-agentcore

TypeScriptMCPAWS BedrockMulti-CloudAI Agents

github.com/agent-dispatch →

02Creator

Askable

A toolkit that gives LLMs structured, real-time UI context — so AI agents can understand what users are looking at and interact more effectively with web applications.

TypeScriptMCPAI AgentsLLM

github.com/askable-ui/askable →

03Creator

claude-harness-builder

A GAN-inspired adversarial multi-agent system for Claude Code — separates Builder and Evaluator agents behind an information firewall so neither can self-approve flawed work.

Claude CodeMulti-AgentShellAI

github.com/vamgan/claude-harness-builder →

04Contributor

Strands Agents SDK

Contributed concurrent message reading for multi-agent sessions, OpenTelemetry cache-usage metrics, corrected tool-result span labeling, and optional outputSchema support for structured tool outputs.

PythonOpenTelemetryMulti-AgentAWS

github.com/strands-agents/sdk-python →

05Creator

Roastmaster

An MCP server that reviews your code and roasts it — identifies bugs and delivers humorous feedback in Gen Z or senior-dev personas. Can auto-fix issues with sarcastic PR comments.

TypeScriptMCPNode.jsClaude

github.com/vamgan/roastmaster →