AI Agent Development
Ram Bikkina builds multi-agent systems for production — supervisor-led delegation, custom MCP tools, and eval hooks — not one-off prompt demos.
Problems this solves
- Single LLM prompts that break on real workflows (IDV, OCR, research, ops)
- Agents that cannot call your internal APIs reliably
- Prototypes that never reach a deployable repo
What can be built
- CrewAI or LangGraph supervisor / specialist graphs
- Custom MCP tools with JSON Schema and structured outputs
- FastAPI backends with SSE streams for agent UIs
- Evaluation hooks for tool-call reliability
Technologies
Python, CrewAI, LangGraph, LangChain, MCP, FastAPI, OpenAI and other LLM APIs, Docker, cloud deploys on AWS/GCP/Azure.
Related work
Engagement
Typical fit: a 2–6 week sprint for one agent loop, or fractional embedding for ongoing agent work. Start a project.