Skip to content
Your-own-LLM
Technology

Built With Modern Technology

We combine modern AI infrastructure, open-source frameworks and full-stack engineering to build practical intelligent systems.

  • AI / LLM

    Frontier and local language models, retrieval and vector search.

    • OpenAI
    • Anthropic
    • Local LLMs
    • RAG
    • Embeddings
    • Vector databases

  • AI Agents

    Orchestration, tool use and autonomous multi-step workflows.

    • CrewAI
    • Browser Use
    • Agent orchestration
    • Tool calling
    • Autonomous workflows

  • Web

    Typed, server-rendered interfaces built for speed and clarity.

    • React
    • Next.js
    • TypeScript
    • Node.js

  • Backend

    Services, data models and APIs that keep systems dependable.

    • Python
    • FastAPI
    • PostgreSQL
    • REST APIs

  • Automation

    Crawling, browser control and connections between existing tools.

    • Firecrawl
    • Browser automation
    • Workflow automation
    • API integrations

  • Voice AI

    Low-latency speech pipelines for real-time conversation.

    • Pipecat
    • Speech-to-text
    • Text-to-speech
    • Real-time AI

  • Infrastructure

    Reproducible environments, version control and deployment.

    • Docker
    • GitHub
    • Cloud deployment
    • Databases
Open Source Technologies We Explore

Tools and frameworks we evaluate and build with

These are independent open-source projects created and maintained by their own teams and communities. We are not their authors — we study them, test them, and use them where they are the right fit for a build.

  • CrewAIMulti-agent orchestrationDiagram of a goal fanning out to several specialised agents and back to one result
    1. Goal
    2. Agent roles
    3. Task hand-off
    4. Result

    Splits a long workflow across narrow agents with clear hand-offs.

  • FirecrawlWeb crawling & extractionDiagram of web pages being crawled into a structured data grid
    1. URL
    2. Crawl
    3. Extract
    4. Structured data

    Turns live web pages into clean, structured input for a model.

  • PipecatReal-time voice pipelinesDiagram of a voice waveform passing through a processing pipeline and back out
    1. Speech
    2. Transcribe
    3. Reason
    4. Reply

    Keeps voice conversations low-latency and interruptible.

  • Browser UseAgentic browser controlDiagram of an agent cursor tracing a path of clicks in a browser window
    1. Instruction
    2. Plan
    3. Browse & click
    4. Outcome

    Lets an agent operate a real browser to finish a task.

  • ClineAgentic coding assistantDiagram of code changes passing a review checkpoint
    1. Request
    2. Plan
    3. Edit code
    4. Review

    Speeds up implementation while a person stays in review.

  • PostizSocial publishing & schedulingDiagram of draft posts passing an approval gate onto a publishing calendar
    1. Draft
    2. Review
    3. Schedule
    4. Publish

    Adds an approval step before anything goes out to channels.

  • AnythingLLMPrivate RAG workspacesDiagram of documents being indexed and funnelled into a single cited answer
    1. Documents
    2. Index
    3. Retrieve
    4. Cited answer

    Answers questions from private documents with sources attached.

Not sure which stack fits your problem? We'll help you choose.

Start a Project