luca baggi.

talks

slides, code and recordings from the times i got on stage.

beyond the hype: building and evaluating robust rag systems workshop

Evaluation-driven development for RAG: bootstrap an evaluation set, judge answers with an LLM, then improve retrieval with HyDE and metadata, with Fabio Lipreri.

  1. AMLD EPFL 2025code

uncertainty estimation at scale with functime, polars and conformal predictions

Fast prediction intervals for thousands of time series with functime and EnBPI conformal prediction.

  1. PyData London 2024video

embeddings, transformers, rlhf: three key ideas to understand chatgpt

The three ideas behind ChatGPT: word embeddings, the transformer architecture and reinforcement learning from human feedback.

  1. PyCon Italia 2024video
  2. PyCon Lithuania 2024video

build your own ai movies buddy workshop

A hands-on workshop on retrieval-augmented generation: build an AI movie expert from scratch, with Fabio Lipreri.

  1. AMLD EPFL 2024code

rust is easy

Start from Python-like code in a Rust program and fix compiler errors one at a time until it is valid Rust.

  1. PyCon Italia 2023slidesvideocode

35 talks given since 2022, 15 different ones

20221 20239 202413 20256 20266

hover or tap a talk to see every time it was given.

20266 talks

  1. inside the mind of an llmPyData Amsterdam 2026
  2. inside the mind of an llmPyData London 2026
  3. inside the mind of an llmPyCon Italia 2026
  4. inside the mind of an llmPyCon Lithuania 2026
  5. from o1 to deepseek: new scaling laws for llms that reasonPyCon Lithuania 2026
  6. mcp: principles and practicePython Milano

20256 talks

  1. from o1 to deepseek: new scaling laws for llms that reasonCodemotion Milan 2025
  2. from o1 to deepseek: new scaling laws for llms that reasonAI Conf 2025
  3. llm inference arithmetics: the theory behind model servingPyData London 2025
  4. foundational models for time series forecasting: are we there yet?PyCon Italia 2025
  5. llm inference arithmetics: the theory behind model servingPyCon DE & PyData 2025
  6. beyond the hype: building and evaluating robust rag systemsAMLD EPFL 2025

202413 talks

  1. foundational models for time series forecasting: are we there yet?PyData Global 2024
  2. foundational models for time series forecasting: are we there yet?PyData Milano at Spotify
  3. polars: zero to heroDealroom
  4. foundational models for time series forecasting: are we there yet?PyData Paris 2024
  5. polars: zero to heroAgile Lab
  6. bigger models or more data? the new scaling laws for llmsAIConf 2024
  7. uncertainty estimation at scale with functime, polars and conformal predictionsPyData London 2024
  8. embeddings, transformers, rlhf: three key ideas to understand chatgptPyCon Italia 2024
  9. how we used polars to build functimePyCon Italia 2024
  10. embeddings, transformers, rlhf: three key ideas to understand chatgptPyCon Lithuania 2024
  11. how we used polars to build functimePyCon Lithuania 2024
  12. build your own ai movies buddyAMLD EPFL 2024
  13. how we used polars to build functimePython Biella Group

20239 talks

  1. polars: zero to heroPython Biella Group
  2. how we used polars to build functimePyData Global 2023
  3. polars: zero to heroPython Biella Group
  4. how we used polars to build functimeKaggle Days Milan 2023
  5. polars: zero to heroPyCon Portugal 2023
  6. polars: zero to heroPyCon Italia 2023
  7. rust is easyPyCon Italia 2023
  8. use rust in your python packagePython Milano
  9. polars: zero to heroPython Milano

20221 talk

  1. coding project templates with cookiecutterPython Milano

35 talks given in 11 cities and online

Milan 9 talks
  1. mcp: principles and practice
  2. from o1 to deepseek: new scaling laws for llms that reason
  3. from o1 to deepseek: new scaling laws for llms that reason
  4. foundational models for time series forecasting: are we there yet?
  5. bigger models or more data? the new scaling laws for llms
  6. how we used polars to build functime
  7. use rust in your python package
  8. polars: zero to hero
  9. coding project templates with cookiecutter
Vilnius 4 talks
  1. inside the mind of an llm
  2. from o1 to deepseek: new scaling laws for llms that reason
  3. embeddings, transformers, rlhf: three key ideas to understand chatgpt
  4. how we used polars to build functime
Florence 4 talks
  1. embeddings, transformers, rlhf: three key ideas to understand chatgpt
  2. how we used polars to build functime
  3. polars: zero to hero
  4. rust is easy
London 3 talks
  1. inside the mind of an llm
  2. llm inference arithmetics: the theory behind model serving
  3. uncertainty estimation at scale with functime, polars and conformal predictions
Biella 3 talks
  1. how we used polars to build functime
  2. polars: zero to hero
  3. polars: zero to hero
Bologna 2 talks
  1. inside the mind of an llm
  2. foundational models for time series forecasting: are we there yet?
Lausanne 2 talks
  1. beyond the hype: building and evaluating robust rag systems
  2. build your own ai movies buddy
Amsterdam 1 talk
  1. inside the mind of an llm
Darmstadt 1 talk
  1. llm inference arithmetics: the theory behind model serving
Paris 1 talk
  1. foundational models for time series forecasting: are we there yet?
Coimbra 1 talk
  1. polars: zero to hero
online 4 talks
  1. foundational models for time series forecasting: are we there yet?
  2. polars: zero to hero
  3. polars: zero to hero
  4. how we used polars to build functime

hover or tap a city to see the talks given there, or a talk to see everywhere it went.

Milan9 talks

  1. mcp: principles and practicePython Milano
  2. from o1 to deepseek: new scaling laws for llms that reasonCodemotion Milan 2025
  3. from o1 to deepseek: new scaling laws for llms that reasonAI Conf 2025
  4. foundational models for time series forecasting: are we there yet?PyData Milano at Spotify
  5. bigger models or more data? the new scaling laws for llmsAIConf 2024
  6. how we used polars to build functimeKaggle Days Milan 2023
  7. use rust in your python packagePython Milano
  8. polars: zero to heroPython Milano
  9. coding project templates with cookiecutterPython Milano

Vilnius4 talks

  1. inside the mind of an llmPyCon Lithuania 2026
  2. from o1 to deepseek: new scaling laws for llms that reasonPyCon Lithuania 2026
  3. embeddings, transformers, rlhf: three key ideas to understand chatgptPyCon Lithuania 2024
  4. how we used polars to build functimePyCon Lithuania 2024

Florence4 talks

  1. embeddings, transformers, rlhf: three key ideas to understand chatgptPyCon Italia 2024
  2. how we used polars to build functimePyCon Italia 2024
  3. polars: zero to heroPyCon Italia 2023
  4. rust is easyPyCon Italia 2023

London3 talks

  1. inside the mind of an llmPyData London 2026
  2. llm inference arithmetics: the theory behind model servingPyData London 2025
  3. uncertainty estimation at scale with functime, polars and conformal predictionsPyData London 2024

Biella3 talks

  1. how we used polars to build functimePython Biella Group
  2. polars: zero to heroPython Biella Group
  3. polars: zero to heroPython Biella Group

Bologna2 talks

  1. inside the mind of an llmPyCon Italia 2026
  2. foundational models for time series forecasting: are we there yet?PyCon Italia 2025

Lausanne2 talks

  1. beyond the hype: building and evaluating robust rag systemsAMLD EPFL 2025
  2. build your own ai movies buddyAMLD EPFL 2024

Amsterdam1 talk

  1. inside the mind of an llmPyData Amsterdam 2026

Darmstadt1 talk

  1. llm inference arithmetics: the theory behind model servingPyCon DE & PyData 2025

Paris1 talk

  1. foundational models for time series forecasting: are we there yet?PyData Paris 2024

Coimbra1 talk

  1. polars: zero to heroPyCon Portugal 2023

online4 talks

  1. foundational models for time series forecasting: are we there yet?PyData Global 2024
  2. polars: zero to heroDealroom
  3. polars: zero to heroAgile Lab
  4. how we used polars to build functimePyData Global 2023

these are historical talks: examples reflect the tools available when presented.