# Pandas Projects at AI Tinkerers

> Canonical HTML: https://aitinkerers.org/technologies/pandas
> Markdown URL: https://aitinkerers.org/technologies/pandas.md
> Technology record last updated: 2026-02-25T03:41:37Z
> Generated: 2026-09-22T21:42:46Z

Pandas is the core Python library for high-performance data manipulation and analysis: it introduces the DataFrame and Series structures for fast, flexible handling of labeled data.

Pandas is your go-to, open-source Python library for data science, engineered for efficient data manipulation and analysis. It centers on two primary data structures: the two-dimensional DataFrame (like a spreadsheet or SQL table) and the one-dimensional Series. This toolkit handles everything from loading diverse file formats (CSV, Excel, JSON) to complex operations: data cleaning, transformation, statistical analysis, and time series processing. Developed by Wes McKinney in 2008, Pandas builds directly on NumPy and provides an intuitive API, making it the industry standard for turning raw data into actionable insights with minimal code.

- Official technology site: https://pandas.pydata.org
- Public AI Tinkerers demos and talks: 10
- Result page: 1 of 1

## Recent Public Talks and Demos

### [From 1,157 handwritten forms to one geological fact: a 25 m pile-driving limit](https://curitiba.aitinkerers.org/talks/rsvp_nt6LgCapvyQ)

A resumable extraction pipeline that turned 1,157 handwritten pile-driving field bulletins — scanned into 1,400+ messy PDFs — into a clean, georeferenced dataset, and then into engineering knowledge. The arc is the point: reports → data → knowledge. Live, I'll show the raw handwritten forms going in (smudged, non-standard, with voided sheets and unrelated pages mixed in), parallel subagents transcribing each to JSON, and the consolidated 1,157-record CSV driving an interactive report — histograms, box plots, an E×N heatmap, and a 3D scatter of 627 georeferenced piles. Then the payoff: those production-control reports became a technical and geological analysis instrument. Aggregated, the field data no single form could show revealed a consistent practical driving limit of ~25 m for the site (mean driven depth 24 m) — a hard geological fact that makes the next foundation estimate far more assertive. The dataset is a real completed job: 1,000+ pre-cast concrete piles driven over six months for a grain terminal in southern Brazil.

- Event context: AI Tinkerers Curitiba: Encontro de Agosto (no EBANX) — 2026-08-26 — Curitiba
- Public talk page: https://curitiba.aitinkerers.org/talks/rsvp_nt6LgCapvyQ

### [Building an ML Decision Lab for Agriculture: Turning Predictions into Learning](https://lausanne.aitinkerers.org/talks/rsvp_4vwa4V9v2ek)

FarmBuddy is an interactive machine learning decision-support system built with Streamlit and a Random Forest regression pipeline. Users can modify agricultural inputs such as crop type, fertilizer usage, land area, and season, then observe how model predictions change in real time. Beyond prediction, the system includes decision logging, before-and-after scenario comparison, input validation, and a learning summary layer designed to help users understand how machine learning models respond to changing conditions. During the demo, I will show the live application, model inference workflow, session-state architecture, and the decision comparison engine.

- Event context: AI Tinkerers Lausanne June 2026 Meetup — 2026-06-25 — Lausanne
- Public talk page: https://lausanne.aitinkerers.org/talks/rsvp_4vwa4V9v2ek

### [Engenheiro civil + Claude Code: app de terraplenagem 100% determinístico construído por conversa](https://curitiba.aitinkerers.org/talks/rsvp_iwCDvtHEc-s)

KML Earthworks é um app Streamlit que transforma um traçado de estrada de acesso desenhado no Google Earth em estimativa de terraplenagem (perfil longitudinal, volumes de corte/aterro pela fórmula do prismatoide, balanço de massa, diagrama de Bruckner) em segundos — upload de .kml, download de Excel, zero GIS de desktop. Na demo eu vou: (1) desenhar um acesso ao vivo no Google Earth e rodar o app em produção (kml-earthworks.streamlit.app) mostrando o pipeline completo (parse → stationing a cada 20m → enriquecimento de elevação com fallback de API → otimização de grade → volumes → export); (2) abrir o repo no editor e mostrar a estrutura src/ modular, o CLAUDE.md que funciona como guideline persistente para o agente, e rodar os 58 testes ao vivo no terminal; (3) navegar pelos commits para mostrar a evolução notebook → pacote → app por desenvolvimento conversacional com Claude Code, incluindo os pontos onde eu tive que intervir manualmente.

- Event context: AI Tinkerers Curitiba: Evento Inaugural — 2026-06-10 — Curitiba
- Public talk page: https://curitiba.aitinkerers.org/talks/rsvp_iwCDvtHEc-s

### [DISAADE](https://conakry.aitinkerers.org/talks/rsvp_e_NuJRZQs6Q)

DISAADE — Là où la technologie rencontre l’amour de la vie. Couveuse néonatale intelligente connectée, conçue en Guinée pour protéger les bébés prématurés. Protéger les plus fragiles, accompagner les mamans, aider les médecins.

- Event context: AI Tinkerers – Conakry: Inaugural AI Conference @ Afrinov Tech Expo — 2026-02-11 — Conakry
- Public talk page: https://conakry.aitinkerers.org/talks/rsvp_e_NuJRZQs6Q

### [MCP Virtual Try On in C++ (KarloAIMutthiMain)](https://dubai.aitinkerers.org/talks/rsvp_tx1DM2AilnQ)

"KarloAIMutthiMain," the second-place project at the AITinkerers Hackathon, is a Virtual Try-On Streamlit application. Users input desired clothing, and the app performs a vector search across an image database to find matches. A key enhancement is the Model Context Protocol (MCP), offering two paths for vector search: - Couchbase Services: For scalable, robust searches. - Local-First Approach: For rapid, low-latency searches. After the vector search, results are sent to an external Replicate inference API. This API performs style transfer, applying the chosen clothing to an image and returning the transformed result to the Streamlit app.

- Event context: AI Tinkerers Dubai Meetup – June 2025 Demo Day — 2025-06-28 — Dubai
- Public talk page: https://dubai.aitinkerers.org/talks/rsvp_tx1DM2AilnQ

### [“Where Did My Money Go?” - Reconciling Ecomm Sales with LLMs, Not Spreadsheets](https://toronto.aitinkerers.org/talks/rsvp_OYdYkJXeDdA)

Every Shopify store owner has faced this mystery: “I sold $6,400, but only $3,900 showed up in my bank account. What happened and Why?” Well, it is because of Stripe fees, customer returns, chargebacks, withholding..etc. We can spend hours to use spreadsheet and complicated formula to match numbers and figure out the difference (the current way) or there must be a better way! In this demo, I’ll show how we’re building an AI agent that can actually reason through transaction data, trace the money trail, and explain where every dollar went—like a forensic accountant, it's faster and no spreasheet. I’ll show the model live digging into real reconciliation logic, identifying discrepancies, and even suggesting follow-up actions. I have only 1 slide to visualize the problem, and then just raw code, live execution and logs to demo whether an (confused) AI agent can really be an accountant.

- Event context: AI Tinkerers Toronto - April 2025 Meetup at Shopify — 2025-04-24 — Toronto
- Public talk page: https://toronto.aitinkerers.org/talks/rsvp_OYdYkJXeDdA

### [Automating the art of deception for the greater good](https://dublin.aitinkerers.org/talks/rsvp_OljSLlzDSv4)

In this talk, I will demonstrate how AI is revolutionizing phishing lure generation at scale, eliminating the need for manual crafting and HTML expertise while making attacks more targeted and convincing than ever before. I will showcase how my AI-powered lure generator can mass-produce highly customized phishing lures using large language models (LLMs)—automating the entire process from text generation to HTML modification. With just a few inputs, AI can generate realistic, organization-specific phishing emails that mimic legitimate communication, allowing for adaptive and hyper-targeted phishing simulations. Additionally, I will explore how a secondary LLM can modify the HTML structure of emails automatically, ensuring that non-technical users can create polished, professional-looking phishing lures without needing to code. This means: ✅ No HTML knowledge required—AI modifies templates dynamically ✅ Mass generation of phishing lures—scaling deception effortlessly ✅ Extremely targeted attacks—customized based on organization, industry, or individual profiles By the end of this talk, you’ll understand how AI-driven phishing lures are reshaping social engineering tactics, how automation removes traditional barriers like coding expertise, and what this means for both attackers and defenders in cybersecurity.

- Event context: AI Tinkerers - Dublin Event (February) — 2025-02-24 — Dublin
- Public talk page: https://dublin.aitinkerers.org/talks/rsvp_OljSLlzDSv4

### [Preswald: Build fast, real-time dashboards with DuckDB and Python](https://sf.aitinkerers.org/talks/rsvp_crPkMiHL2gE)

This talk explores the technical mechanics of leveraging DuckDB, an in-memory OLAP database, to build real-time dashboards using Preswald, our open-source Python framework. Preswald combines DuckDB’s high-performance query engine with a declarative API, enabling developers to create fast, lightweight, code-first dashboards with full control over their analytics stack. - In-memory query execution: DuckDB operates entirely in-memory, processing complex SQL queries directly on flat data files (CSV, Parquet, or raw streams) without the need for databases, data warehouses, or ETL pipelines. - Real-time data transformation: Learn how DuckDB enables on-the-fly data transformations with SQL-like operations that run natively on your data files, avoiding unnecessary data movement or duplication. - Design trade-offs: Preswald’s declarative Python API tightly integrates with tools like Pandas and Plotly, giving devs more control over data transformation, queries, schedules, and visualization workflows. --- From an engineering standpoint, we decided to build Preswald around DuckDB because it let us rethink how analytics pipelines should work. Traditional setups rely on data warehouses, ETL pipelines, and distributed databases, which add a lot of complexity, latency, and maintenance overhead. These systems often require moving data around constantly, which slows things down and makes them harder to manage. DuckDB stood out because of its in-memory, columnar database engine. It’s optimized for vectorized execution and can run complex SQL queries directly on flat files like CSVs and Parquet without needing a separate database server or moving data to another system. This drastically reduces I/O overhead and eliminates the need for extra infrastructure. We saw this as the foundation for building a faster, more lightweight approach to analytics. To make it easier for developers to use DuckDB, we built Preswald with a declarative Python API. It lets developers define their data workflows without having to write raw SQL, while still taking full advantage of DuckDB’s performance. We designed the API to integrate with familiar tools like Pandas for data manipulation and Plotly for visualizations. Under the hood, Preswald uses DuckDB’s vectorized query engine to handle joins, aggregations, and transformations efficiently. A big challenge was balancing performance and flexibility. DuckDB is incredibly fast in-memory, but it can also process data directly from disk for larger datasets. We built Preswald to handle both cases, allowing small, fast prototypes in memory and scaling up to larger workflows without adding extra dependencies or infrastructure.

- Event context: AI Tinkerers - San Francisco - January 2025 Meetup — 2025-01-30 — San Francisco
- Public talk page: https://sf.aitinkerers.org/talks/rsvp_crPkMiHL2gE

### [PREMIER SPORTS](https://dubai.aitinkerers.org/talks/rsvp_9275Yfq6xKc)

Title: Predicting Football Match Outcomes with AI: My Approach to Premier Sports Description: I'm excited to share my project, PREMIER SPORTS, an AI-powered application designed to predict football match outcomes. By leveraging advanced machine learning algorithms and data analytics, I aim to provide accurate predictions that enhance game analysis for fans, coaches, and analysts alike. My approach combines historical team performance data with real-time player and team statistics to deliver more insightful and actionable information. I'm particularly interested in exploring how my model can be fine-tuned using transfer learning techniques, enabling me to adapt to changing team lineups and player performances. I'll also be showcasing some key features of my tool, including: Real-time data ingestion from various sources (e.g., sports databases, social media) Advanced feature engineering techniques for handling high-dimensional data Model selection and hyperparameter tuning Technical Depth and Focus: To provide a more comprehensive understanding of my approach, I'll be focusing on the following technical aspects: Data Handling: I employ a combination of libraries such as pandas, to carry out feature engineering, for labelling and extracting as much relevant data as possible. Transfer Learning Techniques: I leverage pre-trained language models such as Llama to capture contextual relationships within predicted match data and give insights to the application user. Live Demonstration and Code Walk-through: To enhance the audience's understanding of my approach, I'll be providing a live demonstration that includes: Interactive Data Visualization: I'll use a library like Plotly to create an interactive visualization of my data, allowing the audience to explore different scenarios and see how my model generates predictions. Code Walk-through: My demo will include a code walk-through where I explain the reasoning behind specific design choices, highlighting how AI techniques can be applied to improve the accuracy and interpretability of my model. Generative AI Integration: To further enhance the interpretability of my model, I'll be incorporating generative AI techniques: Text Generation: I'll use Llama to generate text-based outputs (e.g., player profiles, and team statistics) that can provide additional context for predictions. Clarity and Engagement To ensure clarity and engagement during the presentation: Concrete Examples: I'll use concrete examples from real-world football matches together with a web UI dashboard, to illustrate how my model can be applied in practice.

- Event context: AI Tinkerers - Dubai Meetup #3 (November) — 2024-11-10 — Dubai
- Public talk page: https://dubai.aitinkerers.org/talks/rsvp_9275Yfq6xKc

### [Network anomaly detection using quantum machine learning](https://dubai.aitinkerers.org/talks/rsvp_kn1saqknpEI)

In this talk, I'll demonstrate how we can harness quantum machine learning (QML) to detect network anomalies faster and more efficiently than traditional methods. Using Qiskit and the KDD Cup dataset, I'll walk through building a quantum circuit-based classifier that identifies suspicious network patterns. I'll explore how quantum circuits can be used for binary classification, the architecture of our enhanced quantum circuit with entanglement layers, practical techniques for data preprocessing and dimensionality reduction, performance optimization strategies for faster training and a demo of the system detecting network anomalies.

- Event context: AI Tinkerers - Dubai Meetup #3 (November) — 2024-11-10 — Dubai
- Public talk page: https://dubai.aitinkerers.org/talks/rsvp_kn1saqknpEI

## Related Technologies

- [Python](https://aitinkerers.org/technologies/python) ([Markdown](https://aitinkerers.org/technologies/python.md)) — 662 public demos
- [Plotly](https://aitinkerers.org/technologies/plotly) ([Markdown](https://aitinkerers.org/technologies/plotly.md)) — 7 public demos
- [Streamlit](https://aitinkerers.org/technologies/streamlit) ([Markdown](https://aitinkerers.org/technologies/streamlit.md)) — 89 public demos
- [Cloud](https://aitinkerers.org/technologies/cloud) ([Markdown](https://aitinkerers.org/technologies/cloud.md)) — 4 public demos
- [NumPy](https://aitinkerers.org/technologies/numpy) ([Markdown](https://aitinkerers.org/technologies/numpy.md)) — 6 public demos
- [PostgreSQL](https://aitinkerers.org/technologies/postgresql) ([Markdown](https://aitinkerers.org/technologies/postgresql.md)) — 106 public demos
- [scikit-learn](https://aitinkerers.org/technologies/scikit-learn) ([Markdown](https://aitinkerers.org/technologies/scikit-learn.md)) — 84 public demos
- [Brandfetch](https://aitinkerers.org/technologies/brandfetch) ([Markdown](https://aitinkerers.org/technologies/brandfetch.md)) — 1 public demo
- [C++](https://aitinkerers.org/technologies/c) ([Markdown](https://aitinkerers.org/technologies/c.md)) — 10 public demos
- [Claude Code](https://aitinkerers.org/technologies/claude-code) ([Markdown](https://aitinkerers.org/technologies/claude-code.md)) — 216 public demos
- [Claude Sonnet 4](https://aitinkerers.org/technologies/claude-sonnet-4) ([Markdown](https://aitinkerers.org/technologies/claude-sonnet-4.md)) — 18 public demos
- [com cooldown mecânico no cliente xml](https://aitinkerers.org/technologies/com-cooldown-mec-nico-no-cliente-xml) ([Markdown](https://aitinkerers.org/technologies/com-cooldown-mec-nico-no-cliente-xml.md)) — 1 public demo
- [Couchbase](https://aitinkerers.org/technologies/couchbase) ([Markdown](https://aitinkerers.org/technologies/couchbase.md)) — 2 public demos
- [Django](https://aitinkerers.org/technologies/django) ([Markdown](https://aitinkerers.org/technologies/django.md)) — 15 public demos
- [DuckDB](https://aitinkerers.org/technologies/duckdb) ([Markdown](https://aitinkerers.org/technologies/duckdb.md)) — 9 public demos
- [FastAPI](https://aitinkerers.org/technologies/fastapi) ([Markdown](https://aitinkerers.org/technologies/fastapi.md)) — 181 public demos
- [Generative AI](https://aitinkerers.org/technologies/generative-ai) ([Markdown](https://aitinkerers.org/technologies/generative-ai.md)) — 45 public demos
- [GPT-4o](https://aitinkerers.org/technologies/gpt-4o) ([Markdown](https://aitinkerers.org/technologies/gpt-4o.md)) — 57 public demos