AI workflow integration
LLM API, RAG retrieval, Dify-style orchestration, review states, and traceable call logs.
AI workflows that become working tools
I help small teams turn repetitive browser, data, and AI review work into reliable systems with logs, retries, handoff docs, and clear human approval steps.
Services
Useful first milestones, not vague transformation decks.
Voice agent backend
Vapi-style tool calls, calendar and CRM routing, transcript summaries, and reusable workflow exports.
Browser and data automation
Playwright flows, CSV/Excel/PDF cleanup, API ingestion, alerts, and retryable failure paths.
Internal tool delivery
Python/FastAPI adapters, lightweight dashboards, Lark/Feishu integrations, and runbooks.
Proof you can inspect
The demo repository shows compact P0 loops: RAG review, Voice AI routing, and a mortgage-servicing slice with CSV import, event logs, and structured LLM classification.
Packages
Start small, verify quickly, then expand only when the first loop works.
Paid trial
Mock or real P0 workflow, field schema, interface notes, and handoff document.
Workflow MVP
One input source, one output destination, structured logs, retry path, and operating guide.
One-week sprint
Several connected workflows or an internal automation surface with review and monitoring.
Start with one bounded workflow
Share a redacted screen recording, sample file, API document, or process outline. I will scope the smallest paid milestone that proves the system can work.