Autonomous software engineering
Built an internal AI software engineer for engineering tasks across the stack. The work connected model-driven task execution with the tools and application infrastructure needed to carry out development workflows.
User-defined skills framework
Built the infrastructure framework that enables users to create their own skills and use them with their own instances of DeepAgent. This work provided the platform capability for user-defined skills.
Native tool calling
Refactored core LLM infrastructure for native tool calling across providers including OpenAI, Anthropic, and Google. This work supported tool-using agents across different model interfaces.
Speech and multimodal interaction
Implemented end-to-end text-to-speech capabilities for DeepAgent with providers including OpenAI, Hume, ElevenLabs, and Gemini.
Persistent files for AI applications
Designed S3-backed persistent storage for user-uploaded files on AppLLM, with presigned upload and download URLs and multipart uploads supporting files up to 5 TB.