Is this you? As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.
Claim your profile
Articles by Kyle Jones
Google releases Lyria 3.5 model – AI music generation
Google’s latest AI music model pushes generative audio further. Lyria 3.5 gives users the ability to pick genre and style in the Gemini app, new styles and templates are designed to start projects ranging from background music to custom tracks. Lyria 3.5 is also available in Google Flow Music with added features in Google AI Studio for developers. Google is making its boldest breakthrough yet with AI music generation, the time couldn’t be more strategic.
Power Options and Swing Options: Modling Volatile Energy Markets
When Hurricane Ida shut down Louisiana’s power grid in August 2021, electricity prices in neighboring markets spiked to $1,500/MWh. Traders holding call options on power made fortunes — their $8/MWh premiums turned into $1,400/MWh payoffs overnight. Meanwhile, generators locked into fixed-price forward contracts faced catastrophic losses, watching spot prices soar 1,800% above their contracted rates.
Why the Three-Tier App Still Matters and How Databricks Simplifies the Data Tier
The three-tier application architecture (presentation, application, and data layers) has been the basic design of enterprise software for decades. It is simple. Each layer has a distinct role. The presentation tier handles user interfaces. The application tier manages logic. The data tier stores and retrieves information. No muss. No fuss. This separation improves security. Each tier can be hardened and governed on its own. It improves resiliency. One tier can fail without collapsing the whole system.
Automate Excel Chart Making with Python
Excel remains the world’s most common business tool. Analysts use it for reports, charts, and dashboards every day. But if you’ve ever updated the same report every week, copied the same formulas into multiple files, or redrawn the same chart for different regions, you know how painful and repetitive Excel can become. Python changes that.
Working with Excel Data in Python
Excel is one of the most common tools for analysts. It’s powerful, but cleaning data in Excel often involves endless formulas, manual edits, and a lot of clicking around. That process can be error-prone and hard to repeat. Python offers a different path. With the Openpyxl library, you can treat Excel workbooks as programmatic objects — read them, transform them with Pandas, and save the results back into a new sheet or file. This approach is faster, reproducible, and easier to maintain.
Stop Wrestling with Excel: Use Python to Automate It
Most analysts live in Excel. It’s familiar, visual, and flexible. But anyone who has pushed Excel past a few hundred thousand rows, chained together too many s, or debugged broken VBA knows its limits. This is where Python shines. With libraries like Pandas and Openpyxl, you can treat Excel files as inputs and outputs in a larger, repeatable workflow. Instead of clicking and dragging formulas by hand, you can automate them.
Why Learning Programming Still Matters in the Age of ChatGPT
Large language models like ChatGPT can now write code that works. You can ask for a Python script to clean data, build a dashboard, or even train a neural network, and in seconds you have something usable. Many people ask a fair question: if ChatGPT can code, why should anyone still learn programming? The answer is that programming is not only about typing lines of code. Programming is about thinking. It is about clarity, logic, and the ability to turn an idea into a system that works reliably.
Pipeline Safety Reporting: Moving the Annual Report Deadline from March 15 to June 15
In June 2025, the Pipeline and Hazardous Materials Safety Administration (PHMSA) announced a proposal to move the annual report filing deadline for gas pipelines and underground natural gas storage from March 15 to June 15, aligning gas operators with hazardous liquid pipeline requirements. This regulatory change aims to reduce compliance costs and relieve unnecessary operational pressure without compromising safety.
Reservoir Geomodeling using Kriging, Geostatistics, and Deep Learning with Python
Transforming subsurface modeling with advanced analytics to reduce uncertainty and accelerate oilfield decisions. Reservoir geomodeling has long depended on interpolation methods that fail to capture complex spatial variability. By integrating Gaussian Process Regression (GPR) and Kriging with a unified analytics platform, operators can now build richer models of permeability and porosity directly from operational and seismic data.
Diagnosing Residual Phase in Seismic Traces Using Python and the Hilbert Transform
Seismic attributes offer a window into subsurface structure, when the wavelet is under control. In many exploration datasets, interpreters rely on visual coherence, amplitude strength, or envelope consistency to track reflectors and make drilling decisions. But subtle errors in wavelet phase can distort structure, push peaks off target, and cloud our picture of the earth.
894 Wax Rd SE, Silver Creek, GA 30173, US
Details Analysis Customize Copied Copied $48,000 894 Wax Rd SE, Silver Creek, GA 30173 Checked: 16 hours ago Updated: Jul 23, 2025 at 06:32PM Investment Summary Net Operating Income (NOI) minus mortgage payments. Calculation: NOI - Mortgage Payments Monthly Cash Flow $567 Capitalization Rate is a rate of return that compares the yearly Net Operating Income (NOI) to the market value.
Structural Break in U.S.
The U.S. housing market shifted in 2020. Listings that once lingered on the market began to vanish in days. Median days on market dropped and stayed low. This is obviously due to COVID, right? Or maybe interest rates? Or something else? Realtor.com tracks monthly medians of “days on market” (DOM), listing prices, inventory counts, and more for the US. Our goal is to understand whether the market changed suddenly in March 2020 and what might explain that shift.
Predicting Pipeline Burst Conditions Using Machine Learning on OpenSees Simulation Data (PVT)
Physical pipelines can burst due to internal pressure. The conditions of bursting can be studied in advance using simulation. This project takes a simulated dataset of pipe response under temperature and pressure variations, and trains machine learning models to predict burst outcomes. We used a dataset manually generated in OpenSees and shared on Kaggle. This article shows how to turn that dataset into a working classification system to predict pipe burst conditions.
Box-Jenkins for Multivariate Time Series Analysis
The Box-Jenkins methodology is a systematic approach to identifying, estimating, and diagnosing time series models. While it is traditionally applied to univariate time series, it can also be extended to multivariate time series using models like Vector Autoregressive (VAR), Vector Moving Average (VMA), and Vector Autoregressive Moving Average (VARMA). This article applies the methodology to two U.S. economic indicators: the Industrial Production Index (INDPRO) and Retail Sales (RSAFS).
Time Series Forecasting Solar Irradiance with ARIMA, SARIMAX, and Gradient Boosting Using SolarAnywhere® Data
Short-term solar irradiance forecasting plays a central role in energy system operations. Utilities, grid operators, and solar developers depend on accurate predictions of solar power availability to balance supply and demand, schedule reserves, and minimize curtailment. One of the most useful quantities in this context is Global Horizontal Irradiance (GHI) — a measure of the total solar energy received per unit area on a horizontal surface.
Crude Pipeline Blending Optimization with Pyomo
Blending crude oil is not a guessing game. Midstream companies do it to meet contractual specs and maximize margin. Each barrel blended below spec costs money. Each unnecessary premium barrel blended in gives away value. A well-built optimization model ensures the right blend goes into the pipe every day. This article walks through a working example of crude blending using Pyomo in Python. We use API gravity and sulfur as constraints. The goal is to minimize cost while meeting product specs.
Portfolio Management in Quant Finance
Portfolio management balances risk and reward. You want to grow capital while protecting it. That tradeoff defines investing. This chapter introduces the essential tools and models used to measure return, manage risk, and optimize the mix of assets in a portfolio. It builds from the basics — risk and return — to advanced concepts like alpha, beta, and the efficient frontier. The return on an asset is the percentage change in value over time.
Time Series in Operations Research
Operations research revolves around making optimal decisions in complex, dynamic systems. Time adds another layer to that complexity. Business environments shift. Demand rises and falls. Resources become constrained. These changes rarely follow a static pattern. Time series analysis addresses that reality. It gives analysts the tools to anticipate shifts, adjust plans, and improve outcomes.
Foundations of Time Series in Forestry (Part 1 of 4)
Forest management requires continuous observation. Forests change slowly, but those changes matter. Decisions about logging, conservation, fire prevention, and biodiversity hinge on understanding how forests evolve over time. Time series analysis provides the structure for that understanding. It reveals trends, cycles, and disruptions in data that might otherwise seem chaotic. This makes it possible to track forest health, predict threats, and plan sustainable resource use.
Why surging demand and grid transformation will define the utility sector in 2025
Kyle Jones is head of solutions architecture at Amazon Web Services. America’s electric power industry is under pressure unlike anything in its history. Surging demand, disruptive technology and climate volatility are converging to reshape how utilities operate, plan and invest. The old playbook no longer works. For utilities to remain relevant — and reliable — they must adapt quickly, creatively and strategically. Nowhere is the shift more visible than in the explosive growth of electricity demand.
The Post-Madison Social Network: Correspondence Between John Adams, 1817-1826
After the final term of James Madison, the American experiment continued — in letters. From 1817 to his death in 1826, John Adams remained one of the most active voices in the country’s political and philosophical life. He wrote. A lot. Thanks to the Founders Online Corpus, we can trace that activity, message by message. This dataset collects and standardizes over 185,000 documents from key American founders. Each record includes metadata like sender, recipient, date, and full text.
Factorial Designs for Causal Inference in Econometric Analysis of Public Policy
Public policy interventions often include multiple interacting components. A job training program might combine skills workshops, mentorship, and placement support. A safety initiative might pair policing strategies with community outreach. Policymakers need to know not just whether each component works, but how they work in combination. Factorial designs offer a method for testing these combinations systematically.
Tracking Houston’s Green Cover with Satellite NDVI using Remote Sensing and Python
Change often happens in ways we can’t quite see — until we pull back far enough. From high above Earth, satellites watch as cities grow, shift, and breathe. Thanks to the European Space Agency’s Sentinel-2 satellites and freely available Copernicus data, we can now track subtle changes in urban green cover, one pixel at a time. This post is about what happens when we look at Houston’s vegetation over the course of 2021, using NDVI — the normalized difference vegetation index.
Differences in opioid prescriptions between subjects with and without intellectual and developmental disability during the COVID-19 pandemic
ABSTRACT Objective This study compared opioid prescription rates between individuals with and without intellectual and developmental disability (IDD) in Utah from 2019 to 2021. Methods Using the 2019–2021 Utah All Payer Claims Database (APCD), we analyzed adults aged 18–62 in 2019, matching those with IDD to those without. We examined opioid prescription frequency, total days of supply, and average morphine milligram equivalents (MME) per prescription each year.
Gas prices don’t change how much Americans Drive. But unemployment does.
Every time gas prices spike, headlines follow. “Will Americans Drive Less?” “High Prices at the Pump Could Curb Travel!” It sounds reasonable. If something costs more, people should consume less of it. That’s basic economics. But reality often ignores our models. We wanted to test this assumption using data — specifically, whether gas prices actually reduce how much Americans drive.
Regression Discontinuity Analysis on COVID’s Affect on Amtrak Commuting
The COVID-19 pandemic changed how Americans move. It shuttered offices, emptied downtowns, and disrupted every part of our transportation network. But three years later, a key question remains: did the pandemic simply pause travel, or did it cause a permanent shift? To answer that, I turned to a rich and underused dataset: Amtrak ridership from 2005 to 2022, broken down by station. This gave us the chance to look not just at totals, but at patterns.
Finding Hidden Patterns in Amtrak Ridership with Dynamic Time Warping
Amtrak serves hundreds of stations across the United States. Some are massive urban hubs. Others are single-platform stops tucked between grain silos and forests. But which ones move together? Which stations follow similar patterns, even if their ridership levels differ? This project set out to answer that question using Dynamic Time Warping (DTW), a powerful tool for comparing time series that don’t align perfectly.
Engagement Isn’t a Survey Score: Why Culture Lives in Daily Conversations
“We filled out the survey. Nothing changed. So why bother?” That was the quiet verdict from an employee after their company launched a months-long, high-budget engagement initiative. The effort looked good on paper: a modern survey platform, glossy slide decks summarizing results, a well-produced video of the CEO talking about the company’s culture. But once the campaign ended, nothing meaningful happened. No policy shifts. No team conversations. No follow-up. Morale declined. Turnover rose.
How to Sustain Innovation Momentum: The Power of Small Wins and Clear Signals
Companies don’t lose momentum because they run out of ideas. They lose it because they don’t know how to keep moving. McKinsey reckons that more than 90% of executives believe innovation is critical to growth — but fewer than 30% believe their companies are good at it. Why? Not because of bad ideas. Because of bad systems. Every innovation initiative starts with a spark. A pilot gains traction. A team gets room to experiment. The energy is high. People are fired up. The kickoff meeting is full of promise.
The Cost of Inaction: Why Fear of Failure Is Riskier Than Trying Something New
What if the safest decision is actually the most dangerous? “Let’s wait. “Let’s see how things shake out.” It sounds prudent. Strategic. Even responsible. But behind those words often sits something more troubling: a bias toward inaction, framed as wisdom.
What It Takes to Lead: Innovation Culture Is a Leadership Test
“You don’t build an innovative culture by asking for big ideas. You build it by how you respond to small ones.” That’s what one senior engineer told us after watching a colleague pitch a modest improvement — and get shut down. The idea wasn’t off-topic. It just wasn’t on the agenda. The manager said, “That’s not really our focus right now,” and moved on. No one argued. No one followed up. But something shifted in the room. That’s how innovation cultures are made — or broken. Not in company all-hands.
The C.O.O.L. Framework: Behaviors That Separate True Innovators From the Rest
I once watched a junior engineer shift the direction of a stalled project — not with code, but with one unexpected question. We were in a cross-functional meeting where the mood was flat. A new initiative had hit a wall: conflicting priorities, no clear owner, too many moving parts. Everyone was looking at their shoes when she leaned forward and asked, “What do you need from us to make this work on your end?” That question unlocked something. The team started talking again. Momentum returned.
Culture Is a Business Strategy: Why Innovation Starts With HR, Not the C-Suite
Most innovation efforts fail before they begin — not because of bad ideas, but because the culture never stood a chance. The first time I heard an HR leader say, “We don’t do culture — we leave that to the CEO,” I froze. She meant well. But the statement revealed something deeper. Too many companies treat culture as a passive outcome. Something that happens when leadership “sets the tone” and the rest of the org just absorbs it. But culture doesn’t trickle down.
Forecasting Retail Sales with Kolmogorov-Arnold Networks (KANs): Beating ARIMA with Deep Function Approximation
Kyle Jones · Follow 4 min read · -- Kolmogorov-Arnold Networks (KANs) are neural network architectures based on the Kolmogorov-Arnold representation theorem, which states that any multivariate continuous function can be expressed as a sum of continuous univariate functions and an auxiliary function. This makes KANs highly expressive and well-suited for modeling complex, non-linear relationships in time series data. KANs can represent complex, non-linear dependencies between variables.
Dynamic Links Between Gold, Newmont, and GDX Using VAR, IRF, and FEVD in Python
Kyle Jones · Follow 7 min read · -- This project uses vector autoregressive models and Granger causality to examine the relationship between the price of gold futures (GC=F), Newmont Mining Corporation (NEM), and the VanEck Gold Miners ETF (GDX). My hypothesis is that the price of gold causes changes in the stock price for NEM and GDX. I used daily data for the five years from Yahoo Finance. The daily data was too volatile to show anything interesting.
Cultural Trends and Media Analysis Using Time Series and Sentiment Analysis of War Terms from 1930-1950 in the US with Python
Kyle Jones · Follow 4 min read · -- Newspapers reflect how a country feels. They capture panic and patriotism, dread and hope. When wars begin or end, the front page often says more about the mood of the people than the official record ever could. This post uses historical newspaper archives to track sentiment shifts across four key terms — war, army, battle, and peace — from 1930 to 1950.
Predictive Maintenance Modeling Time-to-Failure using Survival Analysis in Python (Turbofan Engines)
Kyle Jones · Follow 6 min read · -- Predicting when a machine will fail is one of the most practical applications of survival analysis. We explore three different approaches to modeling failure risk using the FD003 subset of the NASA C-MAPSS turbofan dataset. Our goal is to estimate survival probability over time using multivariate sensor readings. Each engine runs until failure, so the event is always observed, and survival analysis becomes a natural fit.
Alternative Predictive Maintenance Performance Metrics using Cohen’s Kappa, Wilcoxon T, and Confusion Matrices with Python
Kyle Jones · Follow 6 min read · -- In another article, it built out three models for predicting the health state of a turbofan. In this project, I look at some alternative measures for performance of these models.
Nonparametric Regression using k-NN for Correlated Data in Python (example with labor participation rate)
Kyle Jones · Follow 4 min read · -- In practice, data rarely arrive as clean, independent observations. Instead, we deal with sequences — sensor logs, financial returns, or monthly labor force reports — where one point affects the next. This chapter tackles nonparametric regression in such correlated settings. We’ll focus on stationary sequences and uniform mixing, and show how familiar techniques like kernel smoothing and k-nearest neighbors (k-NN) extend to handle these challenges.
How stealing a slide rule got Francis Phillips banished from England in 1786
Kyle Jones · Follow 3 min read · -- This is a true story. The details are straight from the transcript of the trial. January 27, 1786 was cold. It had been an especially cold and dry winter. Francis Phillips pulled his coat tightly around him and crossed the courtyard to the carpentry shop. Separated from the house by about ten feet, Phillips tried the door of the shop. It was fastened with a bolt. He tried the side door, it was locked fast. He crept to the opposite side of the building.
Using Dummy Variables in Econometric Analysis of Public Policy
Kyle Jones · Follow 3 min read · -- Public policy decisions often involve categorical variables, such as geographic regions, political affiliations, industry sectors, or regulatory statuses. These variables influence economic, environmental, and political outcomes, guiding policymakers in designing effective interventions. However, traditional regression models require numerical inputs, making categorical data challenging to incorporate directly. Dummy variables offer a solution.
Time Series Forecasting with Moirai
Kyle Jones · Follow 2 min read · -- Moirai is a time series language model from Salesforce. Named for the three fates who control destiny, it is easy-to-use and fast (wall time was 2.9 seconds for this project). Moirai is a Masked Encoder-based Universal Time Series Forecasting Transformer trained on the Large-scale Open Time Series Archive (LOTSA) [27 billion observations across nine domains]. The goal of Moirai is to make a generalizable model.
LLMs for Time Series Forecasting
Kyle Jones · Follow 14 min read · -- Time series analysis is used for finance, healthcare, industrial IoT, and many other fields. Statistical methods like ARIMA and Exponential Smoothing have been the primary tools for forecasting for decades. More recently, deep learning models such as Long Short-Term Memory networks (LSTMs) and Transformers have redefined the space — now large language models (LLMs) and changing how we do time series analysis.
Distributed Lag Models in Dynamic Model Time Series
Kyle Jones · Follow 4 min read · -- Time series data often exhibit delayed effects, where past values of a variable influence the present. Dynamic models capture these dependencies by incorporating lagged values of explanatory variables. One important class of dynamic models is the distributed lag model (DLM), which explicitly accounts for how past values affect the present over multiple time steps.
Financial Engineering and Time Series Analytics
Kyle Jones · Follow 3 min read · -- Financial markets generate vast amounts of time-dependent data, from stock prices and trading volumes to macroeconomic indicators and alternative data sources. Understanding these time series is essential for developing quantitative trading strategies, managing risk, and pricing complex financial instruments.
Time Series Visualization with Matplotlib and Plotly
Kyle Jones · Follow 4 min read · -- Visualizing time series data reveals patterns, trends, seasonality, and anomalies that are not immediately visible in raw data. Basic line plots offer a starting point, but advanced visualizations provide deeper insights. This chapter explores intermediate techniques for time series visualization using Matplotlib and Plotly. It focuses on interactive features, customizations, and specialized plots that highlight complex patterns and relationships in the data.
The Marginal Cost of Information: When Is Enough Data Enough for Time Series Analytics
Kyle Jones · Follow 5 min read · -- Information is never free. Whether it is collected through surveys, purchased from a third party, extracted from public records, or generated through experimentation, acquiring data always incurs costs. These costs can be direct, such as paying for proprietary data or running a costly field study, or indirect, such as the time spent cleaning, processing, and analyzing raw information.
Transforming Non-Stationary Time Series with Differencing
Kyle Jones · Follow 3 min read · -- Some time series data follows clear trends or patterns, making it hard to model accurately. Most statistical methods assume stationarity — meaning the average level and variability of the series stay the same over time. But real-world data rarely cooperates. Temperatures rise, economies grow, and demand cycles up and down. If you don’t adjust for these trends, your models will misfire. One of the simplest ways to fix this is differencing.
Time Series Data for Digital Humanities
Kyle Jones · Follow 3 min read · -- Time drives history. In the digital humanities, time series analysis uncovers trends, anomalies, and repeating patterns in historical data. A time series is a sequence of data points collected over time. The order of the data matters. Each observation depends on the ones before it. Time series methods track long-term shifts, such as the rise of “freedom” in literature from 1800 to 2000. Political language changes from election to election.
Model Residuals in Time Series Analysis
Kyle Jones · Follow 4 min read · -- Model residuals tell the story of what a forecasting model missed. A solid residual analysis helps validate the model and reveals areas for improvement. This chapter breaks down how to analyze forecast residuals effectively. Residuals are the gaps between what the model predicts and what actually happens. In a well-functioning time series model, these should behave like white noise — random, uncorrelated, and normally distributed.
Show More
loading
Actions
Is this you?
As a journalist, you can create a free Muck Rack account to customize your profile, list your contact preferences, and upload a portfolio of your best work.Get in touch with Kyle
Contact Kyle, search articles and posts on X, monitor coverage, and track replies from one place.
Learn more about Muck Rack