00.00.00

Shivam Singh

AI Engineer & Autonomous Systems Architect

Building autonomous AI agents that think, plan, and execute.

•From self-evolving memory systems to multi-agent orchestration, I craft AI that operates independently while staying aligned with human intent.
•Currently building SCOUT, NEXUS, and experimental agent frameworks.

Get in touch

Here are my socials

Open Source Contributor

Cross-ecosystem maintainer · velero, ray-project, kubeflow, kubernetes-sigs, Cloud-CV, modelcontextprotocol

2025 - Present

GitHub

Merged pull requests across major open-source projects including velero-io, ray-project, Kubeflow, kubernetes-sigs/headlamp, Cloud-CV/EvalAI, the MCP Python SDK, ZenML, and RISC-V infrastructure
Shipped reliability fixes such as routing orphaned backup deletion through DeleteBackupRequest in velero and hiding GPU/GRAM columns when absent in the Ray dashboard
Fixed hard bugs others skipped: MCP respond()/cancel() concurrent-cancellation race, Headlamp events table crash on missing source, and LogReg edge cases in EvalAI
Improving DX and docs across ZenML GitBook rendering, JSON-Schema test-suite spec links, and Python SDK guards

TypeScriptPythonGoRustKubernetesDockerGit

SCOUT

Self-Evolving Memory Agent

2025 - Present

GitHub

Built an AI agent with autonomous memory management that learns, recalls, and evolves its knowledge base over time without human intervention
Designed a persistent memory layer using LangChain and ChromaDB that stores, retrieves, and compacts long-horizon context across sessions
Wired retrieval-augmented generation (RAG) so the agent grounds answers in its own evolving memory rather than static context
Exposed a FastAPI service for memory ingestion, retrieval, and diagnostics with live agent feedback loops

PythonLangChainChromaDBRAGGPT-4FastAPI

NEXUS

Multi-Agent Orchestration Framework

2025 - Present

GitHub

Built a framework for coordinating multiple AI agents to solve complex tasks through collaborative planning and execution
Implemented role-specialized agents that plan, delegate, and verify subtasks using AutoGen and CrewAI
Added an event-driven runtime with Redis pub/sub and WebSockets for real-time inter-agent communication
Enabled conversational handoffs so a coding assistant can spawn, monitor, and merge parallel agent work

PythonAutoGenCrewAIRedisWebSockets

VERITAS

Truth Verification Engine

2025 - Present

GitHub

Built an AI-powered fact-checking system that cross-references multiple sources to verify claims in real-time
Plumbed a RAG pipeline over GPT-4 and BERT embeddings to retrieve supporting or contradicting evidence for a claim
Modeled entity and source relationships in Neo4j to surface provenance and confidence paths
Delivered verdicts with source citations through a FastAPI endpoint

PythonGPT-4BERTRAGFastAPINeo4j

JARVIS

Local AI Assistant

2025 - Present

GitHub

Built a fully local AI assistant integrating LLMs, voice interaction, and automation tools for privacy-first task execution
Wired Whisper for voice transcription and local model inference so queries never leave the machine
Automated system and workflow tasks through a FastAPI service with tool-calling

PythonLLMsWhisperFastAPILocal Inference