Data science/ML/AI
Open in Telegram
Data science and machine learning hub Python, SQL, stats, ML, deep learning, projects, PDFs, roadmaps and AI resources. For beginners, data scientists and ML engineers π https://rebrand.ly/bigdatachannels DMCA: @disclosure_bds Contact: @mldatascientist
Show more
Network:π Analytical overview of Telegram channel Data science/ML/AI
Channel Data science/ML/AI (@datascience_bds) in the English language segment is an active participant. Currently, the community unites 13 926 subscribers, ranking 8 885 in the Technologies & Applications category and 28 496 in the India region.
π Audience metrics and dynamics
Since its creation on Π½Π΅Π²ΡΠ΄ΠΎΠΌΠΎ, the project has demonstrated rapid growth, gathering an audience of 13 926 subscribers.
According to the latest data from 15 September, 2026, the channel demonstrates stable activity. Although there has been a change in the number of participants by 25 over the last 30 days and by 4 over the last 24 hours, overall reach remains high.
- Verification status: Not verified
- Engagement rate (ER): The average audience engagement rate is 7.07%. Within the first 24 hours after publication, content typically collects 2.05% reactions from the total number of subscribers.
- Post reach: On average, each post receives 985 views. Within the first day, a publication typically gains 285 views.
- Reactions and interaction: The audience actively supports content: the average number of reactions per post is 5.
- Thematic interests: Content is focused on key topics such as panda, learning, row, api, ethic.
π Description and content policy
The author describes the resource as a platform for expressing subjective opinions:
βData science and machine learning hub
Python, SQL, stats, ML, deep learning, projects, PDFs, roadmaps and AI resources.
For beginners, data scientists and ML engineers
π https://rebrand.ly/bigdatachannels
DMCA: @disclosure_bds
Contact: @mldatasci...β
Thanks to the high frequency of updates (latest data received on 16 September, 2026), the channel maintains relevance and a high level of publication reach. Analytics show that the audience actively interacts with content, making it an important point of influence in the Technologies & Applications category.
13 926
Subscribers
+424 hours
+37 days
+2530 days
Data loading in progress...
Similar Channels
Tags Cloud
Incoming and Outgoing Mentions
---
---
---
---
---
---
Attracting Subscribers
September '26
September '26
+58
in 0 channels
August '26
+168
in 1 channels
Get PRO
July '26
+119
in 0 channels
Get PRO
June '26
+199
in 1 channels
Get PRO
May '26
+177
in 0 channels
Get PRO
April '26
+277
in 1 channels
Get PRO
March '26
+138
in 1 channels
Get PRO
February '26
+175
in 0 channels
Get PRO
January '26
+171
in 9 channels
Get PRO
December '25
+118
in 1 channels
Get PRO
November '25
+111
in 1 channels
Get PRO
October '25
+181
in 1 channels
Get PRO
September '25
+275
in 2 channels
Get PRO
August '25
+436
in 0 channels
Get PRO
July '25
+312
in 0 channels
Get PRO
June '25
+191
in 1 channels
Get PRO
May '25
+183
in 0 channels
Get PRO
April '25
+233
in 0 channels
Get PRO
March '25
+241
in 1 channels
Get PRO
February '25
+274
in 1 channels
Get PRO
January '25
+765
in 3 channels
Get PRO
December '24
+743
in 1 channels
Get PRO
November '24
+352
in 2 channels
Get PRO
October '24
+328
in 2 channels
Get PRO
September '24
+351
in 3 channels
Get PRO
August '24
+341
in 5 channels
Get PRO
July '24
+383
in 1 channels
Get PRO
June '24
+436
in 1 channels
Get PRO
May '24
+452
in 2 channels
Get PRO
April '24
+522
in 3 channels
Get PRO
March '24
+512
in 5 channels
Get PRO
February '24
+517
in 3 channels
Get PRO
January '24
+511
in 1 channels
Get PRO
December '23
+471
in 0 channels
Get PRO
November '23
+70
in 2 channels
Get PRO
October '23
+87
in 4 channels
Get PRO
September '23
+102
in 0 channels
Get PRO
August '23
+179
in 0 channels
Get PRO
July '23
+132
in 0 channels
Get PRO
June '23
+190
in 0 channels
Get PRO
May '23
+158
in 0 channels
Get PRO
April '23
+129
in 0 channels
Get PRO
March '23
+155
in 0 channels
Get PRO
February '23
+114
in 0 channels
Get PRO
January '23
+181
in 0 channels
Get PRO
December '22
+197
in 0 channels
Get PRO
November '22
+123
in 0 channels
Get PRO
October '22
+244
in 0 channels
Get PRO
September '22
+274
in 0 channels
Get PRO
August '22
+93
in 0 channels
Get PRO
July '22
+81
in 0 channels
Get PRO
June '22
+100
in 0 channels
Get PRO
May '22
+101
in 0 channels
Get PRO
April '22
+160
in 0 channels
Get PRO
March '22
+578
in 0 channels
Get PRO
February '22
+186
in 0 channels
Get PRO
January '22
+129
in 0 channels
Get PRO
December '21
+31
in 0 channels
Get PRO
November '21
+47
in 0 channels
Get PRO
October '21
+28
in 0 channels
Get PRO
September '21
+286
in 0 channels
Get PRO
August '21
+191
in 0 channels
Get PRO
July '21
+252
in 0 channels
Get PRO
June '21
+1 000
in 0 channels
| Date | Subscriber Growth | Mentions | Channels | |
| 15 September | +5 | |||
| 14 September | +1 | |||
| 13 September | +9 | |||
| 12 September | +7 | |||
| 11 September | +2 | |||
| 10 September | +2 | |||
| 09 September | +2 | |||
| 08 September | +9 | |||
| 07 September | +6 | |||
| 06 September | +2 | |||
| 05 September | 0 | |||
| 04 September | +5 | |||
| 03 September | +2 | |||
| 02 September | +5 | |||
| 01 September | +1 |
Channel Posts
| 2 | π What Iβm learning for 2027
Iβve been working in software and data science for over 8 years, but lately Iβd be lying if I said I wasnβt a little worried about where our jobs are heading. π
The future feels more uncertain than ever, so Iβve been thinking seriously about whatβs actually worth learning to stay relevant in 2027 and beyond.
I searched around for resources Iβd personally want to invest my time in, and i figured why not sharing with you guys as well. This is my shortlist π
π§ 1. Letβs Build GPT from Scratch, Andrej Karpathy
Build a GPT yourself and finally understand whatβs happening behind the API.
β±οΈ ~2h
π https://www.youtube.com/playlist?list=PLAqhIrjkxbuWI23v9cThsA9GvCAUhRvKZ
π₯ 2. Neural Networks: Zero to Hero, Andrej Karpathy
A deeper dive into neural networks, backpropagation, language models, GPT and tokenization.
β±οΈ ~19h
π https://karpathy.ai/zero-to-hero.html
π€ 3. Hugging Face AI Agents Course
Learn how AI agents actually work: tools, actions, reasoning and agentic workflows.
π° Free
π https://huggingface.co/learn/agents-course/unit0/introduction
π 4. Designing Data-Intensive Applications, Martin Kleppmann
The classic for understanding databases, distributed systems, replication, partitioning, streams and designing systems that scale.
π ~600 pages
π https://github.com/aasthas2022/SDE-Interview-and-Prep-Roadmap/blob/main/System%20Design/Resources/Designing%20Data%20Intensive%20Applications%20by%20Martin%20Kleppmann.pdf
βοΈ 5. Made With ML
The production side of ML: deployment, testing, monitoring, data pipelines and MLOps.
π° Free
π https://madewithml.com/#course
π― Why these?
My bet for 2027 is that writing code itself will become easier, while understanding AI + production systems + architecture will become even more valuable.
So thatβs what Iβm focusing on.
If you know a resource that belongs on this list please share it so everybody can find it valuable.
Hope this helps β€οΈ | 299 |
| 3 |  Difference Between AI Systems: A Human Analogy | 421 |
| 4 |  βοΈ SQL JOIN Explained Visually
#SQL | 489 |
| 5 | Which isolation level allows a transaction to read data another concurrent transaction has written but not yet committed? | 530 |
| 6 | Machine Learning Visualized
This is an interactive curriculum with animations and exercises that show how machine learning algorithms actually work. You can watch gradient descent, decision boundaries, neural networks, clustering, and more evolve step by step. It is great for building intuition instead of treating models as black boxes.
π¬ Free Interactive + Animation Course
β° Duration: Self-paced
πββοΈ Self Paced
π¨βπ« Created by: Daniel Sobrado / community project
π Link
#MachineLearning #Interactive #Visualization #Course
ββββββββββββββ
π Join @bigdataspecialist for more π | 565 |
| 7 |  5 LLM quantization techniques, clearly explained:
1. RTN: ignores them. Rounds every weight to the nearest grid level with no calibration data. Cheapest option, weakest at low bit widths.
2. GPTQ: repairs after rounding. Quantizes a layer column by column and adjusts the remaining weights to absorb the error before moving on.
3. AWQ: protects before rounding. Finds the ~1% of weight channels that matter most and scales them up so they survive quantization. Everything still ends up in plain INT4.
4. LLM. int8(): isolates at inference. Outlier dimensions run in FP16, the other 99.9% run in INT8, and the results are merged.
5. QAT: solves it during training. The model is fine-tuned with rounding baked into every forward pass, so it adapts to the damage before quantization is actually applied.
All five produce the same artifact, a model at a fraction of its trained precision. They differ only in where the outlier problem gets addressed.
The visual above nicely summarises these techniques.
#LLM | 617 |
| 8 | Chest X-Ray Classification Project
#book | 676 |
| 9 |  π 10 Websites for Finding Real-World Datasets
Finding good datasets is often harder than building the model. These websites cover almost every domain imaginable.
1. Kaggle Datasets
2. Hugging Face Datasets
3. Google Dataset Search
4. UCI Machine Learning Repository
5. OpenML
6. Our World in Data
7. World Bank Open Data
8. data.gov
9. FiveThirtyEight Data
10. AWS Registry of Open Data
You'll rarely run out of project ideas with these bookmarked.
#Datasets | 800 |
| 10 | Local RAG on Windows
#book | 888 |
| 11 |  π² What Makes Random Forest "Random"?
A Random Forest isn't just "many decision trees."
Each tree sees a different random sample of the data.
Then... At every split... It only considers a random subset of features.
So instead of producing 100 identical trees... You get 100 different opinions.
The final prediction is the majority vote (classification) or average (regression).
The randomness is exactly what makes the forest stronger. | 1 044 |
| 12 | SQL Cheat Sheet.pdf | 1 125 |
| 13 | π¦ Your CSV Might Be Using Twice the Memory It Needs
Open a CSV in Pandas.
Run:
df.info()
You'll often notice many text columns have the type:
object
If a column contains repeated values like:
London
London
London
Paris
Paris
Berlin
convert it to:
category
Instead of storing the full text every time, Pandas stores each unique value once and references it internally.
On large datasets, memory usage can drop dramatically. | 1 267 |
| 14 |  NumPy Cheat Sheet for Beginners | 1 136 |
| 15 | From Zero to Data Scientist
This is a free, open-source curriculum from Microsoft's Azure Cloud Advocates team that breaks data science down into 20 digestible lessons spread across 10 weeks.
π Free curriculum with quizzes and assignments
π No prior experience needed to start
π It's Project-based so you're building a portfolio as you learn
π Created by Microsoft experts and students
π Has a strong discord community to back you up
Explore it here: https://github.com/microsoft/data-science-for-beginners | 1 190 |
| 16 |  𧩠Why One-Hot Encoding Exists
Machine learning models don't understand words.
They understand numbers.
So how do you feed a value like:
Color = Red
You can't simply write:
Red = 1
Blue = 2
Green = 3
The model might think Green > Blue > Red, even though colors have no natural order.
Instead, we create separate columns:
Red 1 0 0
Blue 0 1 0
Green 0 0 1
This is called One-Hot Encoding. It represents categories without introducing fake relationships. | 1 011 |
| 17 |  π¨ UC Berkeley just open-sourced FreeToken.
It claims 2β4Γ faster local LLM inference than Ollama, and the wild part is the models it can run:
β’ Qwen3.6-35B on 8GB VRAM β 39.3 tok/s
β’ DeepSeek-V4-Flash 284B on 32GB VRAM β 22 tok/s
β’ GLM-5.2 753B on 96GB VRAM β 14.9 tok/s
How? These are Mixture-of-Experts models. A 35B model doesn't actually use all 35B parameters for every token.
FreeToken keeps the experts in system RAM and intelligently decides whether a missing expert should be sent to the GPU or computed on the CPU.
The good part is the best strategy depends on your exact machine. A 5090 desktop and an 8GB laptop may want completely opposite approaches.
It also checkpoints agent context, so coding agents don't repeatedly prefill thousands of unchanged tokens.
Open weights don't mean much if nobody can afford the hardware to run them. FreeToken is attacking that gap.
π Paper: https://arxiv.org/pdf/2608.16157
π» Repo: https://github.com/FlashML-org/FreeToken | 991 |
| 18 | What does this query return? | 956 |
| 19 | Topic: SQL
π Quick look before the question:
SELECT e.name, e.salary
FROM employees e
WHERE e.salary > (
SELECT AVG(salary)
FROM employees
WHERE department = e.department
); | 925 |
| 20 | πΌ One Pandas Function That Can Save You From Ugly if/else
Suppose you want to classify customers:
spending >= 1000 β VIP
spending >= 500 β Regular
otherwise β Low
You could write a complicated function.
Or:
import numpy as np
df["segment"] = np.select(
[
df["spending"] >= 1000,
df["spending"] >= 500
],
[
"VIP",
"Regular"
],
default="Low"
)
Now the rules are visible directly in the code.
This becomes especially useful when you have several conditions. | 1 033 |
