Hey devs đź‘‹
If you’re building agentic systems (RAG, Voice, copilots, chat agents, Text2SQL, etc.), you’ve probably hit some of these:
“It works on the eval set, melts down on real users.”
“Logs show nothing obvious, but the agent clearly did something dumb.”
“We can’t tell why it picked that tool / branch / answer.”
So for December, we’re running a weekly series:
Fix your Agent - AMA with Future AGI’s engineering team
What it is
Live, open office hours with our Senior Applied Scientist (Rishav) and ML Engineer (Kartik) where we walk through your problems, not slides.
We’ll cover things like:
- Agent debugging & failure analysis
- How to design evals & metrics for agents (not just single LLM calls)
- Prompt optimization strategies that are actually measurable
- Agent observability: traces, decision paths, loop detection
- Architecture trade-offs for production systems (latency, cost, reliability)
Who it’s for
- Backend / ML / data engineers shipping agentic features
- Product folks responsible for reliability and UX
- Anyone trying to move from “demo” to “production” with agents
đź—“ When: Every Wednesday in December
🕤 Time: 9:30 AM PT
đź“Ť Where: Zoom (via Luma)
đź”— RSVP link: https://luma.com/rekjbyfc
Come with:
- A short description of your setup (stack, provider, agent type)
- One or two specific failure cases or questions
- Any logs / traces / sample conversations you can share (sanitized)
We’ll try to cover as many real examples as possible and share patterns that others can reuse.
If you’re planning to join, fill the form so we can prep for your questions and prioritize -> https://forms.gle/gbUZgeFbVsTccVoj8
Top comments (0)