You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Rod Johnson, Embabel. Creator of Spring and Embabel
4
+
5
+
brief: >
6
+
7
+
Ideas for the flow:
8
+
9
+
Gen AI is amazing. show cool examples for personal use, personal assistant use case
10
+
Show failure statistics, recent MIT gartner etc and with quotes
11
+
Diagnose causes for this: Technology is nondeterministic, it's slow and expensive to run, things change all the time, best practice is emerging: ALSO it's not integrated with bsuiness, it's coming out of a diffferent team, Python frameworks aren't aligned with business, it's excessively vendor driven, vendors selling miraclesy
12
+
13
+
How do we fix this? Firstly we attack the nondeterminism and make agents as predictable as possible. Second we seek ways to integrate with what already works and not imagine we can reinvent the world
14
+
15
+
What is the role of Java developers?
16
+
17
+
We could imitate the Python world and copy their approaches. That approach is doomed to failure and if that's the best we can do (a) I'm very worried about the future and (b) go learn Python (you should know it anyway).
18
+
19
+
Or we could do more than play catchup. We could bring our assets to a far superior approach
20
+
21
+
The future of Gen AI for business is substantially on the JVM.
22
+
23
+
Embabel is a framework that attacks every one of these challenges
24
+
- Unique innovative planning element that is smart but deterministic. Actions/Goals allow an extensible system better than FSM of LangGraph or Crew AI but will always do the same thing
25
+
- Emphasis on strong typing and tooling
26
+
- Brings structure to LLM interactions vs just text (consider Embabel vs Python Crew AI examples)
27
+
- Ability to mix code actions and LLM invocation seamlessly for the maximum predictability possible (because of structure)
28
+
- Emphasis on domain modelling and potential to integrate with what's already there
29
+
- Builds on Spring for easy adoption in Java
30
+
- Lovely modern API in Java or Kotlin
31
+
- Highly testable
32
+
33
+
Once more the Java community can do with Spring and open source did in
34
+
taking back control from vendors and demonstrating what works
35
+
36
+
As an example of Embabel's power, this slide deck was built by Embabel.
37
+
The Decker agent illustrates the core ideas. Pull example of how
38
+
Agents and Goals and AchievesGoal work.
39
+
40
+
Conclusion:
41
+
42
+
You could go with vendor snake oil or start from scratch
43
+
with an unimaginative python framework. Massive risk.
44
+
Or try Embabel, the best agent framework everywhere.
45
+
Put a little Gen AI into your apps today and then grow towards the full potential of unlocking your business value
46
+
47
+
header: |
48
+
---
49
+
marp: true
50
+
title: Gen AI Grows Up: Building Production-Ready Agents on the JVM
51
+
theme: default
52
+
paginate: false
53
+
class: invert
54
+
size: 16:9
55
+
---
56
+
57
+
images:
58
+
embabel:
59
+
useWhen: start of deck on title slide, end of deck. it's a tall image
0 commit comments