After a 2 month hiatus I'm moving from the "what" to the "how." pivoting from data scienceto more industrial systems with a focus on SRE, data engineering and the Financial Markets. tads v2.0 is live. 🛠️📈

If you are not putting it out there, you are doing it wrong.

too low yet too aktiv #BuildWithAILagos2026 #BWAI26

TaiResume is live 🚀 tairesume.tadstech.dev AI-powered resume tailoring in seconds. 🎁 5 free coins on signup (limited time)

The way the market gonna reverse on you😂

Prompt engineer decided to learn git

I'm baaaack, who misses me.🙂 Now to get serious...

10X on GOLD✅

Everyone’s posting love. I built something for the people barely holding it together. Check it out and share🙂 sequelove.vercel.app #ValentinesDay2026 #selflove

Sometimes you look back and say what the hell just happened. Then you sit to recalculate and try to integrate not deviate from the desolate yet compassionate solution. Sad.

Step away from the keyboard. The solution will find you. 🚶

D12 of 30 days of TSA. Fitted Autoregressive (AR) models to the first-differenced gold prices. PACF suggested possible order around 4, but AIC and BIC both selected AR(10) as optimal. Coefficients were small, indicating weak autocorrelation. More info: tadstech.web.app/challenge/t…

D9 of 30 D of TSA. Implemented SES on the differenced (stationary) gold price returns. SES applies exponentially decaying weights with optimal alpha around 0.18. Evaluated with RMSE, MAE, MAPE. Challenge tadstech.web.app/challenge/t… #30DaysOfTimeSeries

D8 of 30D of TSA. Explored three moving average techniques on gold prices: SMA, WMA, and EMA. Compared their lag, smoothing effectiveness and responsiveness. more info on my page: tadstech.web.app/challenge/t… #30DaysOfTimeSeries

D7 of my 30 days of TSA. I conducted thorough EDA on the 10-year gold price data and built an interactive Plotly dashboard with gold-themed styling. more info on my portfolio: tadstech.web.app/challenge/t… #30DaysOfTimeSeries

D6. After differencing the gold price data, I analyzed ACF and PACF correlograms to determine ARIMA parameters. Both plots showed no significant lags beyond zero, indicating the differenced series is white noise. This suggests an ARIMA(0,1,0) model tadstech.web.app/challenge/t…

Google is finally deploying the native integration of Gemini 3 into Gmail and it represents a fundamental shift in how we interact with our own data. For the last twenty years email has essentially been a chronological list of files that you had to manually sort and keyword search. With this update Google is effectively turning your inbox into a vector database that you can query in natural language. This is not just a chatbot slapped onto the side bar it is a complete restructuring of the inbox driven by Retrieval Augmented Generation or RAG on your private data. The most significant architectural change is the new search capability for AI Premium users. Instead of typing keywords and getting a list of twenty matching emails you can now ask semantic questions like "What was the Italian restaurant we went to last year in New York?" The system performs a retrieval step over your historical email corpus, identifies the specific reservation confirmation, and generates a precise answer citing the source. This turns the inbox from a storage locker into an oracle where the value is not the messages themselves but the information trapped inside them. This move also signals the start of Agentic Email. The new experimental AI Inbox feature breaks the tyranny of the chronological feed. Instead of sorting mail by time received it uses semantic understanding to group messages by intent. It identifies high priority items that require action and bubbles them to the top while summarizing lower priority updates into digestible clusters. This creates a semantic sorting layer where the LLM acts as an executive assistant that pre reads your mail and decides what actually matters before you even open the app. On the generative side features like Help Me Write are moving from simple auto complete to context aware drafting. Because the model has access to the entire thread history it can draft replies that actually reference previous details rather than just generating generic pleasantries. We are seeing the transition from Smart Reply which was a simple classification task to true generative drafting where the model understands the state of the conversation. By making these features native Google is betting that the killer app for LLMs isn't a new website but the complete transformation of the legacy tools we use every day.

Day 3-5 of my 30 days of TSA. I did not post the last two days due to heavy workload but I am back stronger on day 5 covering the ADF test to statistically confirm non-stationarity, then applied first-order differencing to achieve stationarity. tadstech.web.app/challenge/t… #30DoTSA

Day 2/30 of my Time Series Analysis challenge: Wrestling with the $GLD from yfinance. Today’s win: Moving from raw data to a clean, filtered interactive chart for 2024–2026. Check out the progress: tadstech.web.app/challenge/t… #30DaysOfTSA #Python #BuildInPublic #DataScience

Hi everyone, I am Mike, an applied ML analyst, Math Major and aspiring researcher. Needing to rev up my quite new skills for the new year starting 30d of Time series analysis and building projects along the way Day 1: Fundamentals of TSA tadstech.web.app/challenge/t…