🌟 How LLMs Can Accelerate Data Science

## 🌟 How LLMs Can Accelerate Data Science

[Eric J. Ma](https://ericmjl.github.io/)

`ericmjl.github.io/how-llms-can-accelerate-data-science/`

---

## 💁🏻‍♂️ About me

- 7 years of GPT-4 prompt engineering experience. 
- Proficient in telepathy, can communicate with future technologies. 
- Proficient in sending time-traveling emails.

*I jest*

---

## 💁🏻‍♂️ About me (take 2)

> "Sr. Principal Data Scientist, DSAI (Research) @  Moderna"

- Sr. Principal Data Scientist: Data science team lead 
- DSAI: Computer nerd 
- Research: Science nerd

---

## 🚀 Our team exists to...

> ...make discovery science run at the speed of thought and to quantify the previously unquantified.

---

## 2️⃣ 🧵

---

## 🧵 Thread 1️⃣: LLMs are incredibly empowering

----

### 🤖 Generative AI @ Moderna

- Fully embraced with democratized access to GPT4-128K. 
- Internal academy builds community. 
- Flowering of use citizen-generated use cases.

----

### 🧠 LLMs are

- your most knowledgeable drafting agent 
- prone to mistakes

*Trust but verify!*

---

## 🧵 Thread 2️⃣: Data science necessarily involves writing software

----

### 📈 Data science requires coding

----

### 🧑‍💻 Coding requires software skills

[![Data Science Software Skills](images/data-science-software-skills.webp)](https://ericmjl.github.io/blog/2024/4/5/how-to-grow-software-development-skills-in-a-data-science-team/)

----

### 🖥️ Software skills can be grown

---

## 2️⃣ 🧵 (recap)

- LLMs are incredibly empowering
- Data science necessarily involves writing software

---

## ✨ LLMs Enable DS Teams to Draft Great Software

----

### 📝 Draft New Code

```python
from sklearn.ensemble import RandomForestRegressor
import pandas as pd

# Read the file `data.csv` and create a Random Forest
# regressor model to predict `output` from the rest
# of the columns.
```

The more specific, the better the LLM response.

----

### 📚 Draft Docstrings

```python
import pandera as pa
from .schemas import raw_data_schema

@pa.check_input(raw_data_schema)
def preprocess(df: pd.DataFrame):
    """Preprocess dataframe.

:param df: pandas DataFrame
    """
    # Copilot is usually great at drafting out docstrings
```

----

### 🧪 Draft Tests

```python
@prompt
def draft_tests(function_body):
    """I have the following function

Draft me three unit tests.
    """
```

---

## 💪 LLMs Gives Your Data Science Team Superpowers

----

### 📓 Refactor entire notebooks

```python
@prompt
def refactor(notebook_json):
    """Given the following Jupyter notebook code:

Help me propose a suite of functions
    that can be refactored out from the notebook.
    """
```

----

### 🔍 Debug Stack Trace

```python
@prompt
def debug(stack_trace):
    """Given the following stack trace:

Help me debug what's going on here
    by proposing hypotheses as to why
    I am encountering this error.
    """
```

----

### 📝 Draft Long-Form Documentation

```python
@prompt
def draft_documentation(
    presentation_transcript,
    diataxis_spec,
    documentation_type
):
    """I have the following meeting transcript:

Here is the Diataxis spec for {{ documentation_type }}:

Please draft for me the {{ documentation_type }} using
    the transcript as source material.
    """
```

Check out the Diataxis framework [here](https://diataxis.fr)!

----

## ✍️ Write commit messages

```text
commit 5ffff99bae4611bdcea09187d839e5b40b9ef630
Author: Eric Ma <ericmajinglong@gmail.com>
Date:   Mon Apr 15 15:55:18 2024 -0400

feat(index): update professional title
    and add about sections

- Update the professional title to
      "Senior Principal Data Scientist" at Moderna.
    - Add a humorous "About me" section
      with claims of GPT-4 prompt engineering experience,
      telepathy, and time-traveling emails.
    - Introduce a more serious "About me (take 2)" section
      detailing roles and expertise.
    - Highlight involvement in Generative AI at Moderna,
      emphasizing democratized access to GPT,
      community building through an internal academy,
      and the development of citizen-generated use cases.
```

----

## ✍️ If you want the tool...

```bash
pip install -U llamabot
llamabot git install-commit-message-hook
# Ensure you have an OpenAI API Key
```

----

### 🗃️ Draft Database Queries

```python
@prompt
def nl2sql(nl_request, schema):
    """Given the following database schema:

And the following natural language request:

Return for me the SQL necessary to satisfy the
    natural language request.
    """
```

----

### ❓ Ask Questions of Code Repos

```bash
llamabot repo chat https://github.com/webpro/reveal-md.git \
--checkout main \
--model-name gpt-4-0125-preview \
--panel
```

----

----

### 📚 Learn a New Domain

```python
@prompt
def learn(concept, education_level):
    """You are an expert in {{ concept }}.

I am looking to understand {{ concept }}.

Explain it to me the way Richard Feynman would explain it
    to a student in {{ education_level }}.
    """
```

----

----

### 😃 Add Emojis to Your Talk

```python
@prompt
def add_emojis(talk_md):
    """Given the following Markdown slide deck:

Add emojis to all of the title headers
    while preserving the original text.
    """
```

----

## 🧩 The Hardest Routine Parts of Data Science

1. Taking the time to document our work. 
2. Checking correctness of our work. 
3. Rapidly ramping up into a new knowledge domain.

*LLMs can accelerate all of the above.*

---

## 🚀 A Productive New World

- LLMs are an insanely great productivity tool. 
- Supercharge your learning and automate the boring stuff.

---

## 🤔 Select reading

- [Grow DS team software development skills](https://ericmjl.github.io/blog/2024/4/5/how-to-grow-software-development-skills-in-a-data-science-team/)
- [Survey of LLM tooling](https://ericmjl.github.io/blog/2024/2/1/an-incomplete-and-opinionated-survey-of-llm-tooling/)
- [Organize and motivate a DS team](https://ericmjl.github.io/blog/2024/3/23/how-to-organize-and-motivate-a-biotech-data-science-team/)
- [Keep technically sharp as a DS team lead](https://ericmjl.github.io/blog/2024/2/25/how-to-keep-sharp-with-technical-skills-as-a-data-science-team-lead/)
- Data Science Bootstrap Notes [website](https://ericmjl.github.io/data-science-bootstrap-notes/get-bootstrapped-on-your-data-science-projects/) and [ebook](https://leanpub.com/dsbootstrap/)

---

## ⭐️ Thank You