Giorgio Robino

Conversational AI technical lead at www.almawave.it

Working now on healthcare multimodal assistants. In past, as researcher at ITD-CNR, I built CPIAbot.

Principal interests:
- Chatbots
- Conversational Agents
- Personal Voice/Multimodal Assistants
- Voicebots and voice interfaces
- Realtime (Industrial) voice-cobots

2025-02-02

My latest paper introduces Conversation Routines—foundational natural language instructions that enable Conversational Agentic Systems to execute business logic workflows. In my upcoming follow-up paper, I’m exploring a CR evaluation framework leveraging prompt engineering.

arxiv.org/abs/2501.11613

2025-02-01

🚀Minor update to my latest paper:
Conversation Routines: A Prompt Engineering Framework for Task-Oriented Dialog Systems

Updated today on arXiv:
arxiv.org/abs/2501.11613v2

#promptEngineering #LLMs #AgenticAutomation #AIAgents #ConversationalAI

2025-01-09

I just published my new article:

Conversation Routines: A Prompt Engineering Framework for Task-Oriented Dialog Systems

convcomp.it/conversation-routi

🧵Exploring a no-code framework for designing task-oriented conversational agentic systems (CAS) using Conversation Routines (CR)!

2024-10-24

SWARMing Conversational AI
Integrating No-Code and Code in Agent-Based Workflows

My last article:
linkedin.com/pulse/swarming-co

2024-07-16

Check out my latest article: Testing the Language Proficiency of Popular LLMs linkedin.com/pulse/testing-lan via @LinkedIn

Giorgio Robino boosted:
2024-06-21
2024-06-11
2023-09-16

From a practical perspective, there is much we can learn from David's work in prompt design. The basic examples he has shared on the GitHub repository demonstrate a refined approach to crafting effective (LLMs) prompts (e.g. in splendid markdown format).

2023-09-16

I view David's work from two different perspectives: he is one of the few individuals with a clearly defined vision regarding what a cognitive architecture should encompass, including all the components of the stack, from the foundational LLMs to the uppermost and critically important Aspirational Layer.

By the way, Dave published three books about cognitive architectures, available for free in their digital version:
lnkd.in/dvjsykKF

2023-09-16

I recommend that all prompt engineers and individuals interested in cognitive architecture (design) consider David Shapiro's new open project: the Autonomous Cognitive Entity (ACE) framework:
lnkd.in/dDfdVswB

2023-09-16

During my current holiday break, I created Prompter.vim — an exciting Vim editor plugin that allows you to run Azure OpenAI and/or OpenAI Large Language Model completions right within the editor.

This effectively turns your editor into both an LLM playground and an efficient editing system, all in one package. I believe this tool can be valuable to fellow prompt engineers like myself.

github.com/solyarisoftware/pro

Giorgio Robino boosted:
2023-09-12
2023-09-11

Question for #LLMs prompt engineers (anyone):

What are the most important metrics for measuring LLM completion?

I start with some basics "system" variables:

1. Latency (or response time)
measures the elapsed time of a completion

2. Tokens Throughput = tokens/latency

3. Prompt/Completion tokens ratio

4. there are many more cross ratios maybe useful, involving LLM settings (max_tokens, temperature, etc.)

The final goal is to define some list of common vars to evaluate foundational LLMs.

2023-08-11

Last but not least, these Google tutorials are interesting, but... Has Google not yet released any LLM (e.g., Bard)? Am I mistaken?

I would greatly appreciate any thoughts or suggestions on how to implement an LLM-based agent that requires multiple tools."

2023-08-11

"LLMs' 'function calling' is a crucial area of interest for me, envisioning task-oriented goal conversational AGENTs involving autonomy APIs calls (referred to as 'tools' below).

youtube.com/watch?v=P3buv6P_u7

2023-02-09

é ora possibile contribuire con annotazioni in lingua italiana su #openassitant 🇮🇹🙌:

open-assistant.io/it/dashboard

VIDEO su come funziona:

youtube.com/watch?v=64Izfm24FK

ANNOTATE CON QUALITA' E FATE GIRARE A PALLA!

#chatGPT #opendata #FOSS #crowdsourced #italiano #linguaitaliana

2023-02-08

#openassitant è il progetto!

E' on-line da oggi la versione in italiano! 🇮🇹🙌:

open-assistant.io/it/dashboard

VIDEO introduttivo su come funziona il meccanismo di annotazione: youtube.com/watch?v=64Izfm24FK

PARTECIPATE E FATE GIRARE!

#chatGPT #opensource #crowdsourced #language #italiano

Giorgio Robino boosted:
2023-02-08

L'Italiano è una delle lingue in cui è possibile addestrare Open Assistant, il progetto collaborativo (guidato da LAION) di assistente OPEN ispirato a #ChatGPT

#OpenAssistant

Grazie a @solyarisoftware per la segnalazione!

Cruscotto dell'interfaccia di addestramento di open-assistant.io

Client Info

Server: https://mastodon.social
Version: 2025.04
Repository: https://github.com/cyevgeniy/lmst