Jason L.

I have focused on AI agents since 2022, especially systems that use tools, retrieve and process documents, run longer workflows, generate structured reports, automate browser or desktop tasks, and interact over voice. Most of my work is beyond the basic chatbot stage. I build custom agent orchestration and integrations, work with existing frameworks such as LangGraph when appropriate, and debug systems that work in a demo but run into problems with real data, batches, concurrency, latency, or production edge cases. Recent projects have included a medical document-review workflow with multi-stage retrieval, financial analysis and presentation generation for commercial real estate, computer-use agents, and a local GPU/SIP voice-agent stack for outbound calls. I created an open-source, plugin-driven agent framework called MindRoot, which I use as a development platform and as a way to build reusable integrations. Clients do not need to adopt it. I am comfortable working in an existing Python or JavaScript stack, using another agent framework, or extracting only the relevant components into a standalone client-owned project. I have experience with voice agents, including an end-to-end voice stack from agent harness to local device inference (STT-LLM-TTS) to SIP (so it does not require providers for inference).