I share the code, explanations, and discussions behind my work. This page brings those resources together so you can learn from them, explore the projects, and contribute improvements.
I founded AgenticWorks to help developers understand AI systems and build reliable agents.
| Resource | What you can use | Code or supporting material |
|---|---|---|
| NeuralNet Fundamentals | A ten-part learning track covering the path from matrices to neural networks, with worked examples and knowledge checks. | Notebook repository |
| Google ADK deep dive | An illustrated guide to agent types, tools, orchestration, state, evaluation, and deployment. | Examples and references appear in the article. |
| AI Trainings | Three Gemini, vector-search, and RAG labs with sample data, reference solutions, and a dated validation record. | Lab notebooks and documentation in the repository. |
The AgenticWorks learning index distinguishes the available NeuralNet track from Agentic Systems and Context RAG tracks being authored. Browse the learning index.
Silos to Synergy: Architecting Scalable Multi-Agent Systems
I presented lessons about data readiness, system-level evaluation, orchestration, and permission boundaries. The associated AI-mmunity community-health project uses Google ADK, Gemini, BigQuery, and Cloud Run.
- Presentation photos, project, and team
- Technical recap and public discussion
- Architecture case study and my contribution: I led the team, integrated teammates' code, and set up agent orchestration.
Our AI-mmunity team won Google Cloud's Agentic AI Arena. The team page credits Abhi Ram Salammagari, Eyoha Girma, Semaa Amin, Aashna Kunkolienker, Tianchen Cai, and Sreekanth Kannan. Team account · Google's event page.
I'm an avid participant in the Bay Area hackathon community and have been part of multiple winning teams. I also enjoy hosting hackathons with companies and local developer communities. These events are a chance to build with others, exchange ideas, and learn through working prototypes.
Alongside AI-mmunity, my Agent Master hackathon recap shares another team project and win.
When a better score still deserves a rejection turns a CX Lab policy-design question into a runnable exercise. Compare a baseline and two patches, inspect attempted tool actions, and reproduce a promotion decision. The source, synthetic cases, per-case results, and tests are included.
This is a deterministic teaching exercise, not a live model benchmark or a customer-impact claim.
I use LLM Arena to investigate interactions between language models, including cooperation, competition, persuasion, and goal adherence. The public repository documents experiment controls, judge-based evaluation, and report generation.
My academic research with Georgia Tech's Design & Intelligence Laboratory and AI-ALOE focuses on socially aware AI, learner engagement, and social presence in online classrooms. I hold an M.S. in Computer Science from Georgia Tech and an M.S. in Applied Artificial Intelligence from the University of San Diego.
Recent engineering reflections connect my projects to questions about evaluation, grounding, and useful interfaces. The Next presentation discussion includes exchanges about agent routing and identity across agent calls.
If you try a resource, share the example you used, what you expected, and what happened. That makes it easier to answer questions and improve the material for the next person.