Members-Only
Recent Talks & Demos are for members only
You must be an AI Tinkerers active member to view these talks and demos.
Can a local-first agent on small local models discover and verify general knowledge
Explore MOiRA, a local-first agent using small models to build and verify a knowledge model. Discover its capabilities, state representation, and the limits of 20-30B parameter models.
MOiRA is a WIP research agent to build a knowledge model of a domain to answer a user’s question.
MOiRA is a general-purpose research system that uses local models (20B-30B range) and can automatically acquire new capabilities through API ingestion.
MOiRA generates verified answers to user questions by reasoning about facts and conclusions, with full transparent traceability to sources and a structured knowledge model.
MOiRA is a local-first Python and LangGraph research agent utilizing OpenAPI ingestion and LanceDB semantic tool discovery.
- LangGraphA low-level orchestration framework for building long-running, stateful, and cyclic multi-agent systems using a graph-based architecture.LangGraph is the specialized, low-level runtime for developing complex AI agents, extending the LangChain ecosystem to handle intricate, stateful workflows. It models the agent's logic as a directed graph: nodes represent actions (LLM calls, tool use), and conditional edges dictate the flow, enabling critical features like cycles (loops) for iterative reasoning. This graph-based approach ensures durable execution, allowing agents to persist through failures and resume operations. Key capabilities include comprehensive memory management via a shared state object and built-in human-in-the-loop functionality (interrupts) for external oversight. This robust framework is trusted by production teams at companies like Klarna and Replit for deploying scalable, resilient agent architectures.
- FastAPIFastAPI is a modern, high-performance Python web framework for building APIs with automatic OpenAPI documentation.FastAPI is a robust, high-speed Python web framework: it is built on Starlette (for async capabilities) and Pydantic (for data validation and serialization). Leveraging standard Python 3.8+ type hints, the framework automatically generates interactive API documentation (Swagger UI/ReDoc) and enforces data validation, effectively reducing developer-induced errors by an estimated 40%. This architecture delivers performance on par with Node.js and Go, significantly increasing feature development speed (up to 300% faster). It is production-ready, fully supporting OpenAPI and JSON Schema standards for all API specifications.
- NaiveUIAn exceptionally complete, TypeScript-first Vue 3 component library featuring built-in dark mode and effortless dynamic theming.Naive UI delivers over 90 production-ready Vue 3 components designed to scale without bloating your bundle. Built entirely in TypeScript, it features a unique CSS-in-JS design engine that eliminates the need to import external stylesheets (saving you configuration headaches). The library excels in customization: you can inject a custom theme dynamically via a single component wrapper, making dark mode transitions and brand-specific overrides remarkably straightforward.
- Local Large Language ModelsRun powerful generative AI models directly on your own hardware with complete data privacy and zero API costs.Local Large Language Models (LLMs) shift artificial intelligence from expensive cloud data centers straight to your personal machine or private servers. By utilizing optimized runtimes like Ollama and llama.cpp, you can deploy cutting-edge open models (such as Llama 3.2, Gemma 4, and DeepSeek R1) on standard consumer hardware. This local-first architecture guarantees absolute data privacy, eliminates subscription fees, and allows offline execution. It is the ultimate setup for developers building secure, low-latency AI applications without cloud dependencies.
- API ingestionAPI ingestion automates the extraction and loading of semi-structured data from external endpoints directly into your data warehouse.API ingestion solves the headache of manual data extraction by setting up automated pipelines to query external REST or GraphQL endpoints. Instead of writing custom scripts for every SaaS tool, you configure a central connector to handle rate limits, manage OAuth tokens, and parse JSON payloads. The system extracts raw data (like customer events from Salesforce or marketing metrics from HubSpot) and loads it directly into analytical environments like Snowflake or BigQuery. By automating pagination and schema mapping, it ensures your downstream dashboards always run on fresh, reliable data.
Related talks
More from the community
Deterministic Safety Gates Before The LLM
DC
Learn how to implement deterministic safety gates before LLMs for high-stakes applications, ensuring robust input interpretation and analysis…
Wisdom Layer: Agent Cognitive Architecture
Manchester NH
This talk introduces the Wisdom Layer, an agent architecture providing persistent memory, autonomous reflection, and self-authored behavioral rules…
Building a Stock Market Research Agent with LangGraph
St. Louis
Learn how to build an agent that automates stock market due diligence, pulling real-time data, profiling leadership, and…
Agent based maths knowledge assessment
Dublin
Your AI is Two Versions Behind
Zürich
Learn how to fill AI model knowledge gaps by structuring recent changes into markdown folders. See a Go…
Making a local, persistent and self-curated wiki with the Andrej Karpathy knowledge base
Pereira
Learn to build a self-curated wiki using Markdown and LLMs, a RAG approach without vector databases, for persistent…
Compose Email
Loading recent emails...