I build practical AI products: agent pipelines, local RAG systems, Ollama workflows, automation tools, and production SaaS experiments.
Practical Docker Compose stacks for Ollama: Open WebUI, n8n, Flowise, ChromaDB, Langfuse, Jupyter, code-server, LibreChat, and multi-GPU local AI. Why star it: it solves a clear developer problem: start useful local AI stacks quickly, with security notes, helper scripts, and a contribution path for new stacks. | Document Q&A with Ollama, ChromaDB, FastAPI, Streamlit, source citations, Docker, and private local knowledge-base workflows. Why star it: it gives developers a clean starting point for private document chat and local RAG experiments. |
Python ReAct-style local agents with Ollama tool calling for research, code review, file reading, and data analysis workflows. Why star it: it is small enough to understand and useful enough to fork into real agent workflows. | Ollama-powered Python scripts for document summarization, code review, test generation, entity extraction, email drafting, and batch workflows. Why star it: it is a practical toolbox for local AI tasks people run repeatedly. |
Prompt libraries and Ollama Modelfiles for coding, writing, RAG, analysis, business, creative workflows, and expert personas. Why star it: it helps local LLM users get better outputs without building a whole app first. | Instruction dataset preparation, LoRA training entrypoint, Unsloth workflow notes, and GGUF/Ollama handoff guidance. Why star it: it gives beginners a concrete path through data prep, training setup, and Ollama packaging. |
Production-grade invoice processing through 4 chained Claude reasoning agents: extraction, validation, classification, and approval decision. Why star it: multi-agent architecture, prompt chaining, validation logic, audit trail, and a real business workflow. | Production-ready interview prototype with Next.js, FastAPI, Whisper STT, local or API LLMs, WebSocket sessions, scoring, and report export. Why star it: real-time interview flow, local-first privacy notes, pluggable LLM provider, and Docker Compose setup. |
| Kit | What it gives users | Status |
|---|---|---|
| Ollama Docker Stacks | Docker Compose stacks for Ollama, Open WebUI, n8n, Flowise, ChromaDB, Langfuse, Jupyter, code-server, LibreChat, and multi-GPU local AI. | Public |
| RAG Starter Kit | Local document Q&A with Ollama, ChromaDB, FastAPI, Streamlit, citations, API, and Docker setup. | Public |
| AI Agent Starter Kit | Local research, code review, and data analysis agents with tool calling. | Public |
| Local AI Automation Scripts | Scripts for summarization, code review, extraction, writing, research, and batch workflows. | Public |
| Ollama Prompt Pack | Prompt libraries and Modelfiles for coding, writing, RAG, business, analysis, and expert personas. | Public |
| Fine-Tuning Starter Kit | Dataset preparation, LoRA training entrypoint, GGUF/Ollama handoff notes, and sample instruction datasets. | Public |
I have worked across:
- AI agents for invoices, interviews, research, code review, and data analysis.
- Local AI tooling around Ollama, RAG, prompt packs, model workflows, and Docker deployments.
- SaaS products for education, HR, resume workflows, social posting, programmatic SEO, and operations.
- Observability and database tooling around Postgres monitoring, log analysis, and automation.
- Founder-friendly product kits that can be cloned, run, modified, and shipped.
# Invoice automation pipeline
git clone https://github.com/raja24790/InvoiceOps
cd InvoiceOps
npm install
npm run api
# AI interview prototype
git clone https://github.com/raja24790/ai-interviewer
cd ai-interviewer
docker compose up --build
# Local AI stack
git clone https://github.com/raja24790/ollama-docker-templates
cd ollama-docker-templates
./scripts/start-stack.sh open-webui
- Run InvoiceOps with your own invoice examples and suggest new edge cases.
- Try AI Interviewer locally and improve the scoring/reporting flow.
- Try Ollama Docker Stacks and request the next stack you want added.
- Star the repos that save you time so I know what to build in public next.