Developer Week 055
Books
AI Engineering by Chip Huyen (20/521) ~4%
NOTES: AI Engineering by Chip Huyen: Preface
Work Log
Designed and implemented a new Customer Service Agent Orchestration Workflow, reducing average response times by 47% from ~17 seconds to ~9 seconds, while increasing message coverage by over 10X, and reducing LLM costs by optimizing prompts and proper model selection.
Mon 7/28
- Learned Jira basics and planned out my first two week sprint
- Redesigned the AI Agent Orchestration workflow
- Began the process of prompt refinement
Tue 7/29
- Learned Postman basics for API testing
- Restructured JSON objects for Agent data processing
- Reflowed to allow for sub-agent orchestration.
- Checked data flow to test/verify JSON objects arrive as expected throughout pipeline.
- Created a sku Availability agent
Wed 7/30
- Identified a
thread_ID
bug and fixed - Implemented the new API for calling o4-mini from our own backend.
- Reflowed the Availability Agent and reached functional state.
- Will be ready to turn on and ship within 3 days!
- Finished Postman course and setup tests for our internal APIs
Thu 7/31
- Finalized
order_status_update
Agent workflow - Launched the full workflow (superior to past version)
- Went live with only one minor bug.
- Created a new
kill_request
flow to prevent spam and system messages - Modified prompt to stop
other
as a category - Updating categories based on the historical data provided to me by Matt
Fri 8/1
- Created the order
cancellation_request
agent scaffolding- Received MVP diagram approval from Chris
- Proactively monitoring the live agent, and modifying orchestrator agent to tweak outcomes as needed. (e.g.
kill_request
for FB notifications) - Created
order_complaint
category - Stayed late to fix
order_status
agent bug that I discovered.
Sat 8/2
- logged errors and prepared testing to first thing Monday morning.