S Shivamani Jadi
Currently building CereFlow — open source on GitHub

Shivamani
Jadi.

CS + Data Science  ·  Kutztown University  ·  Junior, 2027

I build agentic AI systems and multi-agent orchestration — and I'm betting my time on the idea that the future belongs to people who can orchestrate AI well, not just the people who write every line by hand.

Résumé GitHub LinkedIn Email
§01About

I study Computer & Information Sciences with a minor in Data Science at Kutztown University (GPA 3.67, four-time Dean's List, Presidential Scholar). My core belief: every AI model is powerful enough to solve almost any problem — the real work is finding the right environment to deploy it in.

That's where I live: designing the architecture, picking the right model for each task, and wiring agents together so the whole system does something a single model can't. I'm a systems thinker first — I understand the concepts deeply and direct AI to execute against them. Most of my hands-on time goes into the frontier of agent tooling: Hermes Agent, OpenClaw, mem0 memory architectures, and multi-model setups across Anthropic, Gemini, and open-source models.

Right now I'm building CereFlow, founding and teaching the Build 2 Earn AI community, and seeking real-world experience in how AI systems become things businesses actually depend on.

§02Selected Work
2026—

CereFlow — Multi-Agent AI Orchestration Framework

A 4-layer agentic AI system I'm architecting to run on a user's own infrastructure — local machine or self-hosted VPS — keeping all data inside an isolated per-user database with no third-party cloud dependency. Layer 1 routes goals to downstream layers that analyze your data and inbound signals (like Gmail), draft a response, ask for your feedback the first time, and act autonomously after that. Built from smaller specialized open-source models rather than one large LLM, with a memory layer modeled on the mem0 architecture.

PythonMulti-Agent OrchestrationAnthropic / GeminiOpen-Source LLMsmem0
GitHub ↗
2025—

Build 2 Earn — AI Training Community

A 20-member community I founded to turn everyday AI users into AI power users. I designed a 20-step, 6-week curriculum from my own research and run weekly live training sessions, continuously folding in emerging tooling — Google Spark & Omni, Hermes Agent, OpenClaw, mem0, Roo Flow — and translating it into workflows people can actually use.

CommunityAI EducationCurriculum DesignLive Training
Details
2024—

SSD Studio — Automated Client Pipeline

Where I learned to ship. I founded a creative-media studio and built its entire automated client lifecycle — cutting booking down to a clean 5-minute self-serve flow, then handling instant confirmations, 2-day and 1-day reminders, day-of notifications, post-shoot follow-up, unedited-image delivery within 12 hours, and the full edited package within 48–72 hours. It taught me that code only matters when it serves a real outcome.

Workflow AutomationReactVercelClient Systems
Live ↗
2025

5-Card Draw Poker Engine & Game UI

A fully functional poker engine with a hardcoded Java backend — deck management, hand evaluation, and a betting state machine — paired with an interactive 2D game interface in JavaFX. I used Google AI Studio's Stitch and Opal to prototype the frontend interaction patterns. Next on the roadmap: a 3D upgrade.

JavaJavaFXGoogle AI StudioStitchOpal
GitHub ↗
§03Stack
AI Systems & Agents
Multi-Agent OrchestrationHermes AgentOpenClaw mem0Prompt EngineeringOpen-Source LLM Config Roo FlowCursor
Models & Platforms
Anthropic ClaudeClaude CodeOpenAI GPT Google GeminiSparkOmni StitchOpal
Languages
PythonJavaJavaScript C++HTML/CSSSQL
Tools & Workflow
JavaFXGit & GitHubVercel LinuxGoogle AI StudioGoogle Workspace
§04Elsewhere
↗
Open Source
CereFlow
My multi-agent AI orchestration framework, in active development.
 ↗
Professional
LinkedIn
Background, communities, and current focus.
 ↗
Creative
SSD Studio
Photography brand & portfolio — portrait, couple, event, fashion.
 ↗
Community
Google Developers
Member, Hudson chapter. Launching a GDSC at Kutztown in Fall 2026.
§05Contact

The fastest way to reach me is email.

Seeking real-world experience in how AI systems become things businesses depend on. Always happy to talk agent architectures, multi-agent orchestration, or AI tooling.