Jayesh Ranghera

Building AI for governance & policy problems
Python · ML · NLP | Political Science + Tech
Learning in public | IBM Certified

[Jayesh Ranghera @jayeshranghera]

70% of AI use cases are non-compliant with the EU AI Act.

I built a free open-source classifier — input any AI system, get instant risk level + compliance checklist.

XGBoost + RAG pipeline on 514 real incidents.

🔗 https://t.co/1gbp0eJMI5

#EUAIAct #AIGovernance

[Jayesh Ranghera @jayeshranghera]

3 questions that define the future of AI governance:
#1 How far can the current paradigm go?
#2 How much can AI improve AI?
#3 Will future AIs still be tools — or something else?

@hlntnr nailing the core uncertainty policymakers must address now.
#EUAIAct #AIGovernance

[Jayesh Ranghera @jayeshranghera]

The EU AI Act is enforcing in 2026.

Most companies have no idea if their AI is compliant.

I'm building a free, open-source compliance classifier — input your AI system, get instant risk level + recommendations.

#EUAIAct #AIGovernance

[Jayesh Ranghera @jayeshranghera]

Is it safe to run outside today in Berlin?

Built a live ML dashboard to answer that.
7-day AQI forecast using 2.37M rows of real EU sensor data.

XGBoost + DuckDB + Streamlit

https://t.co/qnRb5eUT91

#Python #DataScience #AirQuality

[Jayesh Ranghera @jayeshranghera]

182,000 premature deaths from air pollution in Europe. In 2023. From PM2.5 alone.

Built an ML dashboard that forecasts AQI 7 days ahead, compares cities, and checks EU law compliance.

2.37M rows. XGBoost. DuckDB. Streamlit.

https://t.co/qnRb5eUT91

#DataScience #ML #Python

[Jayesh Ranghera @jayeshranghera]

OpenAI's GPT-5.3 Codex helped build itself.

AI debugging AI.
AI evaluating AI.
AI improving AI.

We just crossed a line most people didn't notice.

The builder is now part of the build.

#MachineLearning #AI #OpenAI #DataScience #BuildInPublic

[Jayesh Ranghera @jayeshranghera]

The White House just released a National AI Legislative Framework.

The EU voted on AI Omnibus today.

Two superpowers. Same week. Racing to govern AI.

And India?

Still waiting.

That silence is exactly the gap I'm building for.

#AIGovernance #EUAIAct #DataScience #BuildInPblc

[Jayesh Ranghera @jayeshranghera]

If you are in tech, let's connect 👋

[Jayesh Ranghera @jayeshranghera]

The lesson nobody tells you about data science:

Finding the right data is harder than building the model.

EDA started. Parquet files loaded. NO2 readings from 2022 onwards.

Now building. Updates coming.

#BuildInPublic #DataScience #AirQuality

[Jayesh Ranghera @jayeshranghera]

6 days. 3 data sources. 0 usable data.

That's what building a real data science project actually looks like.

Here's what happened with my Urban AQI project

[Jayesh Ranghera @jayeshranghera]

Tried Kaggle next. Found big datasets but couldn't process them locally.

Then found EEA — European Environment Agency.
Real NO2 pollution data. Parquet format. Multiple stations. Proper timestamps.

Not perfect. But real.

Sometimes the right data is just the data that works.

[Jayesh Ranghera @jayeshranghera]

Attempt 1 — OpenAQ API
Data fetched fine. But location data was wrong on their server side.
Delhi showed 11,800 stations. Most pointed to wrong coordinates.
Scrapped it.

Attempt 2 — WAQI
Data came in. But not enough volume for meaningful ML.
2 APIs. 3 days
Scrapped it too.

[Jayesh Ranghera @jayeshranghera]

AI will not take your job.

But someone using AI better than you will.

The real skill in 2026 isn't coding.

It's knowing WHAT to build and WHY.

Technical skills expire. Problem-solving doesn't.

#MachineLearning #DataScience #AIGovernance #BuildInPublic

[Jayesh Ranghera @jayeshranghera]

US Treasury just launched an AI governance initiative for financial systems.

Finally.

But India has 1.4B people, zero AI regulatory framework, and I'm building models for exactly this gap.

Governance can't be an afterthought.

#AIGovernance #DataScience #BuildInPublic

[Jayesh Ranghera @jayeshranghera]

Day 1 of building Urban Air Quality Prediction dashboard.

Delhi alone: 11,800 air quality monitoring stations.

Stack: OpenAQ API → Databricks → PySpark → Hugging Face Spaces

Just getting started. Updates coming.

#BuildInPublic #DataScience #AirQuality

[Jayesh Ranghera @jayeshranghera]

Built an Android health Tracking app today.

Claude + Gemini + Cursor AI.
Zero manual code. 3 AI tools. 1 working app.

This is how we build in 2026.

#BuildInPublic #AI #Vibecode

[Jayesh Ranghera @jayeshranghera]

97% accuracy sounds great.
Until you realize the model predicts "no fraud" for everything.
What actually matters:
-Recall: did you catch real positives?
-Precision: were predictions correct?
-F1: balance of both
Match your metric to the cost of being wrong.
#MachineLearning #Data

[Jayesh Ranghera @jayeshranghera]

Feel Like Boss😎

[Jayesh Ranghera @jayeshranghera]

Political Science grad → Data Scientist.

No CS degree. No coding background.

Just 1 rule: no tutorial without a real project.

1 year - python to ML models deployed.

You don't need a CS degree to break into data science.
You need curiosity and consistency.

#MachineLearning

[Jayesh Ranghera @jayeshranghera]

Built an ML system that predicts industrial equipment failures before they happen. 96.2% accuracy. 94.8% recall. Streamlit dashboard. €8.5M projected savings. Full breakdown 👇🏻https://t.co/93m773llQI #MachineLearning #DataScience #Python