About SignalForge

Built by an AI Engineer. For Autonomous Systems.

SignalOS is the flagship platform by SignalForge — built by Rithvikh Katpelly, an AI & ML engineer at UC Davis shipping production-hardened RAG, multi-agent, and MCP-native systems.

SignalForge · Engineering autonomous trading infrastructure.

The Builder

Why I built SignalOS

I'm Rithvikh Katpelly — an AI & ML engineer-in-training at UC Davis, pursuing a B.S. in Data Science & Economics (expected June 2028). I build production-hardened RAG and agentic AI systems end-to-end: hybrid retrieval architectures combining structured SQL and unstructured vector data, multi-agent orchestration with evaluation suites gating CI, and tool-calling agents hardened against prompt injection.

SignalOS is the intersection of everything I've been building toward — autonomous pipelines, MCP-native architecture, and risk-first engineering. I've shipped a live MCP server on Google Cloud Run, a multi-agent economic analysis system over FRED, and a production-style support triage system using the OpenAI Agents SDK. Every project is a step toward a coherent, autonomous trading OS.

Outside engineering, I compete as an NCAA student-athlete in men's tennis at UC Davis — which means I know what it takes to perform under pressure, manage time across competing demands, and iterate fast. That mindset is baked into how I build.

Certifications

Claude Certified DeveloperAnthropic Academy
IBM RAG & Agentic AI10-course certificate
AI Agent DeveloperVanderbilt · Coursera
GCA Perfect ScoreCodeSignal · 600/600
RK

Rithvikh Katpelly

Builder, SignalOS · SignalForge

UC Davis · B.S. Data Science / Economics

Technical Stack

Multi-Agent SystemsMCP ProtocolRAG PipelinesLangChain / LangGraphOpenAI Agents SDKpgvector / FAISSGoogle Cloud RunAWS BedrockPyTorch / TensorFlowPython · C++ · SQL

30+

Cert Hours

10+

AI Projects

600

GCA Score

“I don't just build demos — I ship production-hardened systems with evaluation suites, CI/CD, and security boundaries baked in from day one.”

Engineering Principles

How SignalOS is engineered

Six principles drawn from Rithvikh's production AI systems — applied to every layer of the platform.

Multi-Agent Orchestration

SignalOS uses a supervisor-agent pattern routing to specialist agents that fan out concurrently — Data, News, Analysis, and Presentation agents coordinating over FRED and live market sources.

MCP-Native by Design

Every system component exposes MCP tools. Built on FastMCP and deployed on Google Cloud Run, the architecture lets any AI client — Claude, Cursor — inspect state, query decisions, and monitor live pipelines.

Hardened Against Injection

Agents are hardened against prompt injection from untrusted external sources with inter-agent trust boundaries, secret redaction, rate limiting, and audit logging — security documented from day one.

Evaluation-Gated CI

Every pipeline ships with a multi-case evaluation suite scoring tool selection, groundedness, orchestration, and injection resistance. Regressions block CI — quality is not optional.

Hybrid Retrieval Architecture

RAG pipelines combine structured SQL and unstructured vector retrieval (pgvector, FAISS, ChromaDB) with semantic chunking, Hugging Face embeddings, and cross-encoder reranking for precision.

Production-First Deployment

Full CI/CD via GitHub Actions: lint, pytest across Python 3.11/3.12, Docker smoke tests, and keyless Workload Identity Federation deploy to Google Cloud Run with automatic rollback on failure.

Join the private beta.

SignalOS is being built in the open. Join the waitlist and be part of the first cohort of traders and developers who get access.