hai

[Neo Kim @systemdesignone]

If you want to elevate your AI engineering career (in 2026), save these 20 GitHub repositories:

1 OpenClaw
↳ Runs a personal AI agent locally that can browse, plan & take actions on your device.

2 TensorFlow
↳ Provides a production-ready framework to build, train & deploy machine learning models at scale.

3 AutoGPT
↳ Automates multi-step tasks by chaining LLM reasoning into autonomous agents.

4 n8n
↳ Automates workflows with a visual builder that integrates APIs, data & AI tools.

5 Ollama
↳ Runs open LLMs locally with simple commands & optimized performance.

6 Stable Diffusion WebUI
↳ Generates images locally with a powerful UI for Stable Diffusion models.

7 Hugging Face Transformers
↳ Offers thousands of pretrained models for NLP, vision & multimodal AI tasks.

8 Langflow
↳ Builds & tests LLM pipelines visually using a drag-and-drop interface.

9 Dify
↳ Creates production-ready AI apps with built-in orchestration, prompts & APIs.

10 LangChain
↳ Orchestrates LLM workflows, tools, memory & agents in applications.

11 Open WebUI
↳ Delivers a self-hosted ChatGPT-style interface with local & API model support.

12 DeepSeek-V3
↳ Provides a high-performance open-weight LLM optimized for reasoning and coding.

13 PyTorch
↳ Builds & trains deep learning models with flexible, research-friendly APIs.

14 Gemini CLI
↳ Interacts with Google’s Gemini models directly from the command line.

15 llama cpp
↳ Runs LLaMA-style models efficiently on CPUs & local hardware.

16 Whisper
↳ Transcribes & translates speech with high accuracy using deep learning.

17 ComfyUI
↳ Designs advanced image generation workflows using node-based pipelines.

18 CrewAI
↳ Coordinates multiple AI agents to collaborate on complex tasks.

19 RAGFlow
↳ Implements retrieval-augmented generation pipelines for enterprise search & QA.

20 Claude Code
↳ Assists coding with deep repository understanding & agent-style workflows.

What else should make this list?

===

💾 Save & RT to help others get good at AI engineering.
👤 Follow @systemdesignone + turn on notifications.

[Akshay 🚀 @akshay_pachaar]

Train your own LLM from scratch.

This repo builds a GPT-style transformer from the ground up, without using any high-level libraries.

You see exactly how attention, multi-head attention, the feed-forward block, embeddings, residuals, and layer norm fit together.

And it doesn't stop at the model. It walks the whole path from raw data to generated text.

↳ Data download, preprocessing, training, and generation
↳ Training data from The Pile (825GB across 22 sources)
↳ Tokenized with tiktoken (r50k_base) and stored in HDF5
↳ Training loop with eval, LR decay, and crash-safe checkpoints
↳ An SFT and RLHF guide for what comes after pretraining

The same code scales by changing a few config values. Around 13M parameters is where the output starts producing correct grammar and spelling, and you can train that in about a day on a free Colab or Kaggle T4.

If you've ever wanted to actually see how a transformer works instead of importing one, this is a clean place to start.

Link to the repo in the comments.

Interested in ML/AI Engineering? Check my FREE AI engineering Guidebook with 380+ pages (downloaded over 80k times, link below)

[Sumanth @Sumanth_077]

10 GitHub Repositories you should definitely check as an AI Engineer!

1. Hands on AI Engineering

Curated repository of AI-powered applications and agentic systems showcasing practical use cases of LLMs

👉 Check this out: https://t.co/pR5riPhGDO

2. Hands on Large Language Models

This repository contains the complete code examples from the book Hands-On Large Language Models.

It includes notebook examples that cover everything from the introduction to language models to fine-tuning them.

👉 Check this out: https://t.co/TuFg99vK5o

3. AI Agents for Begineers

Beginner friendly course on AI Agents

This Free 11-lesson course will teach you everything you need to get started with building AI agents.

👉 Check this out: https://t.co/UHk9lCk18P

4. GenAI Agents

This repository provides tutorials and implementations for various Generative AI Agent techniques, from basic to advanced.

It serves as a comprehensive guide for building intelligent, interactive AI systems.

👉 Check this out: https://t.co/4c1JL0lZHe

5. Made with ML

Learn how to design, develop, deploy and iterate on production-grade ML applications.

Check this out: https://t.co/TBcsWr1DOi

6. Learn Harness Engineering

A project-based course on building the environment, state management, verification, and control mechanisms that make AI coding agents work reliably.

👉 Check this out: https://t.co/MfqVEA9hiB

7. AutoResearch by Andrej Karpathy

Learn how to build autonomous ML experiment loops where AI agents modify training code, run experiments, and iterate on their own.

This 630-line Python script shows you how to set up an agentic research workflow that runs ~100 experiments overnight on a single GPU. Practical implementation of autonomous research systems.

👉 Check this out: https://t.co/bfdhkX2u17

8. Designing Machine Learning Systems

This repo contains the summaries and resources for Designing Machine Learning Systems book

👉 Check this out: https://t.co/dBbKls3oOr

9. Awesome LLM Inference

Curated list of LLM/VLM inference papers with codes covering Flash-Attention, Paged-Attention, WINT8/4, Parallelism, and more.

Comprehensive resource for LLM inference optimization techniques including quantization, KV cache management, attention mechanisms, and deployment strategies.

👉 Check this out: https://t.co/Dtvkw7uwuR

10. LLM Course: The best hands-on course to learn Large Language Models with roadmaps and Colab notebooks!

👉 Check this out: https://t.co/dNJqw8a7pd

If you found it insightful, reshare with your network.

Find me → @Sumanth_077 for more insights and tutorials on AI Engineering!

[Vincent | 信号＞噪音 @VincentLogic]

一个女孩靠“空气手势控制3D”赚了42万美金！

技术门槛低到离谱：
一个80刀的摄像头 + 两个免费软件（TouchDesigner + MediaPipe）。

最狠的是：
核心代码全是 Claude 帮她从零写出来的。

她靠这个去时装秀和发布会做现场交互表演，一场收费 5000-15000 美金。据说诺兰的新电影也用了类似的技术。

我刚才顺着她的思路用 Claude Code 跑了一遍，一个周末真能复刻出来。

这算不算普通设计师/程序员最直接的搞钱路子？

[rob. @robiartec]

10 repositorios de IA virales en GitHub esta semana:

1. Understand-Anything

Convierte cualquier base de código en un grafo de conocimiento interactivo.

→ Visualiza estructura, dependencias y lógica como un mapa
→ Buscas y preguntas sobre el código
→ Funciona con Claude Code, Cursor, Copilot y Gemini CL

https://t.co/XqjoBs1WSx

2. Codegraph

Indexa el código antes para que los agentes no busquen archivos en cada consulta.

→ -57% tokens y -62% llamadas a herramientas
→ 100% local

https://t.co/VaYb23gEWb

3. MoneyPrinterTurbo

Genera vídeos cortos solo con palabras clave.

→ Crea guion, material, subtítulos y música
→ Vertical 9:16 y horizontal 16:9

https://t.co/sANaCc9dJN

4. ECC

Mejora los entornos multiagente como Claude Code, Cursor y Codex.

→ Habilidades, memoria y seguridad en un paquete
→ 63 agentes y 249 habilidades

https://t.co/2ZOSbNL9W2

5. Taste-skill

Saca a la IA de los diseños UI genéricos.

→ Aplica layout, tipografía y espacios en blanco de nivel pro
→ Funciona con Claude, Cursor y Codex

https://t.co/Sf5wcHfZ6Y

6. knowledge-work-plugins

11 plugins oficiales de Anthropic para convertir Claude en experto por área.

→ Ventas, soporte, marketing, legal y finanzas
→ Pensados para Claude Cowork, también valen en Claude Code

https://t.co/zK3ngg1m07

7. Anthropic-Cybersecurity-Skills

754 habilidades de ciberseguridad para tus agentes.

→ Compatible con 5 frameworks, entre ellos MITRE ATT&CK
→ Ojo: lleva “Anthropic” en el nombre pero es de la comunidad, no oficial

https://t.co/chmTVtpPoo

8. academic-research-skills

Cubre todo el flujo de investigación: papers, revisión y corrección.

→ La IA hace lo tedioso (verificar citas), tú mantienes el criterio
→ Diseño “IA como copiloto”

https://t.co/s2JLf5LZFl

9. pi

Kit completo para desarrollar agentes de IA.

→ CLI de codificación, API unificada para varios LLMs, infra de ejecución y UI de terminal
→ Integra con Slack y blinda la cadena de suministro

https://t.co/gg7DWHmt8y

10.skills (oficial de Anthropic)

Colección oficial de ejemplos de Agent Skills.
→ Escribes el procedimiento en SKILL.md y lo reutilizas en Claude Code, https://t.co/aBpDQ8eV8m y API
→ Plantilla para crear tus propias habilidades

https://t.co/P1nDd6wvkt

Resumen de la semana: grafos de código + habilidades para agentes.

La IA empieza a leer el código como un mapa y a cambiarse de habilidades según lo que necesite.

Si trabajas con agentes, esto es lo que viene.

Guarda esta lista 🔖

[Matt Dancho (Business Science) @mdancho84]

RIP toy projects.

If your portfolio doesn’t touch real business problems, you’ll get filtered out.

Here are 300+ real ML system case studies from top companies (free).

[Avi Chawla @_avichawla]

I have been fine-tuning LLMs for over 2 years now!

Here are the top 15 techniques I'd learn if I were to fine-tune them:

Bookmark this.

1. LoRA
> Freezes the base weights and trains two low-rank matrices as the update, resulting in ~95-99% fewer params to fine-tune.

2. QLoRA
> LoRA on top of a 4-bit quantized base model.

3. Prefix tuning
> Prepends trainable vectors to keys and values at every layer, weights frozen.

4. Adapter tuning
> Inserts small trainable modules between transformer layers.

5. Instruction tuning
> Supervised tuning on (instruction, response) pairs so the model follows directions instead of just continuing text.

6. P-tuning
> Optimizes continuous prompt embeddings through a small encoder, mainly for NLU tasks where discrete prompts are unstable.

7. BitFit
> Trains only the bias terms, ~0.08% of params, and still rivals full fine-tuning on small-to-medium datasets.

8. Soft prompts
> Steer a frozen model with learned vectors instead of handcrafted tokens.

9. RLHF
> Train a reward model on human preference rankings, then PPO against it. The pipeline behind the first ChatGPT.

10. RLAIF
> Swaps the human labeler for an LLM judging. RLHF-level quality at a fraction of the cost.

11. DPO (Direct Preference Optimization)
> Skips the reward model and optimizes preference pairs directly with a classification-style loss. Simpler than PPO.

12. GRPO (Group Relative Policy Optimization)
> Samples a group of responses per prompt and normalizes their rewards within the group. DeepSeek R1 ran on it.

13. RLVR (Reinforcement Learning with Verifiable Rewards)
> Replaces the learned reward model with a checker or compiler returning verifiable scores. The free signal behind R1's math and code.

14. Multi-task fine-tuning
> Trains on several tasks at once, so one model generalizes and shares representations instead of overfitting to one objective.

15. Federated fine-tuning
> Tunes across decentralized clients that share only weight updates, never raw data. For when data can't leave the device.

GRPO needs exactly one scalar reward per response. RLVR (13) produces that for free on math and code by running the answer through a checker or compiler.

But tasks like a RAG answer, a support reply, or a summary have no gold label to match against.

The usual fallback is a hand-written reward function scoring faithfulness, hallucination, and completeness.

It takes days to calibrate, rewards the wrong behavior when the weights are off, and breaks every time you add a tool or edit the system prompt.

RULER, implemented in OpenPipe's ART (open-source), solves this.

During training, it passes the N sampled trajectories to a judge LLM, which ranks them relative to each other against the agent's system prompt and returns the scores.

Relative ranking is more stable than absolute scoring, and GRPO normalizes within the group anyway, so the rankings feed straight into the pipeline like with RLVR.

Here's the GitHub Repo: https://t.co/srt6poinAz

(don't forget to star it ⭐ )

I wrote a full breakdown recently on how exactly this works, with the training loop and code.

Read it below.

[Nico @nicos_ai]

Ahora puedes darle memoria infinita a Claude, Codex y Cursor.

100% gratis y open source.

Memanto ya es tendencia con +5.000 estrellas en GitHub.

La herramienta:
→ guarda el contexto de tus sesiones de trabajo
→ lo comprime y organiza con IA
→ recupera lo relevante en menos de 90ms
→ funciona con Claude Code, Codex, Cursor, LangGraph, CrewAI

Sin bases de datos vectoriales y sin configuración compleja.

Un solo comando para instalarlo:

→ pip install memanto

Ya no reinicias el contexto. Tu agente simplemente lo recuerda.

Enlace abajo👇

[Mr. Buzzoni @polydao]

the Atlassian engineer who was laid off dropped a full guide to becoming a senior engineer after 15 years in the industry

Vasilios Syrakis - no university degree. Made it to Senior Engineer anyway

the things that actually worked:

> taught himself Regex by mass-answering Stack Overflow until he became the expert people came to
> learned Python for free, immediately built a DNS web interface and shipped it - didn't wait to feel ready
> watched the same conference talks dozens of times until concepts actually stuck

if he was starting from zero today:

> get a CS degree - he skipped it and says that was a mistake
> build a home Kubernetes cluster and try to sell what you make
> grind LeetCode - not for the code, for the vocabulary to prompt Claude correctly
> share everything publicly - the audience compounds faster than the skills
> show up to meetups in person

the one mindset shift that separates juniors from seniors: when you join a new team, write a deep analysis of how everything works before touching a single line of code

> 15 years at the top of the industry and he still gets impostor syndrome

so does everyone else who's actually good

the roadmap is in the video 👇

[Elias Al @iam_elias1]

10 free GitHub repos that people shouldn't miss.

Each one replaces something you are currently paying for.

Save this thread. You will come back to it.

🧵 Thread:

[Pratham @Prathkum]

Everything you need to know about REST API:

[Dhruv @dhruvtwt_]

A list of useful tools for web focused design engineers.

Link : https://t.co/voUmLhS9yd

[Dan Kornas @DanKornas]

Stop learning LLMs from disconnected tutorials.

LLM from Scratch is a hands-on PyTorch curriculum for builders who want to understand how LLMs are trained, modernized, and aligned.

It helps you move from concepts to implementation by organizing the path from transformer basics to tiny-model training, scaling, fine-tuning, reward modeling, and RLHF.

Key features:

• End-to-end curriculum – follows pretraining → finetuning → alignment from foundations through RLHF
• Transformer from first principles – covers positional embeddings, self-attention, attention heads, MLPs, residuals, LayerNorm, and full blocks
• Tiny LLM training loop – includes tokenization, batching, cross-entropy, sampling, validation loss, and a no-Trainer training loop
• Modern architecture upgrades – walks through RMSNorm, RoPE, SwiGLU, KV cache, sliding-window attention, and streaming cache ideas
• Alignment path included – covers SFT, reward modeling, PPO-style RLHF, and GRPO with concrete training-loop notes

It’s open-source (GPL-3.0 license).

Link in the reply 👇

[Your Sophia Ai @NextShopiaAi]

LIST OF 40 WEBSITES TO FIND REMOTE JOBS

1. Linkedin. com

2. Indeed. com

3. Glassdoor. com

4. FlexJobs. com

5. weworkremotely. com

6. Remote. com

7. Upwork. com

8. Freelancer. com

9. Fiverr. com

10. Guru. com

11. Toptal. com

12. AngelList. com

13. Hubstafftalent. com

14. Simplyhired. com

15. Remotive. com

16. Virtualvocations. com

17. workingnomads. com

18. Hired. com

19. cloudpeeps. com

20. taskrabbit. com

21. talent. com

22. Remote OK - remoteok. io

23. DRemote - dremote. io

24. Jooble - jooble. org

25. stackoverflow. com/jobs

26. jobspresso. com

27. onlinejobs. ph

28. simplyhired. com

29. themuse. com

30. skipthedrive. com

31. zirtual. com

32. justremote. com

33. hireable. com

34. remoteworkhub. com

35. jobbatical. com

36. freelancewritinggigs. com

37. contentwritingjobs. com

38. problogger. com/jobs

39. behance. net

40. designhill. com

Don't forget to follow @NextShopiaAi to get more insightful Ai related tools and update.

[Noorie lova @noorielova]

Finally cracked the code on this cafe filter!

🍎iphone14pm
•Exposure: -13
•Vibrance: +8
•Highlights: -10
•Shadows: -18
•Brightness: +6
•Black Point: +5
•Sharpness: +28
•Clarity: +18

[Harman @itsharmanjot]

10 WEBSITES THAT FEEL ILLEGAL TO KNOW ABOUT.

Bookmark every single one of these.

1. https://t.co/7SLjuIK4GR
Removes paywalls from almost any news article in 1 click.

2. https://t.co/rm7EPs1dve
A full Photoshop running inside your browser. Opens PSD, AI, Sketch, and XD files.

3. https://t.co/NmzOmyDjOJ
Solves math, physics, finance, and chemistry problems your calculator can't.

4. https://t.co/gvXHjC2Y21
Erases the background from any photo in 3 seconds. No signup.

5. https://t.co/Xqh8s1DQYU
Tracks every plane in the sky in real time. Free, no account needed.

6. https://t.co/PnjritkGBM
Records any meeting and transcribes it word for word with speaker labels.

7. https://t.co/EaZ5iimDNq
Generates a working email address in 1 second. Useful for sketchy signup forms.

8. https://t.co/OtTvIVSCil
The Wayback Machine. Lets you read deleted articles, dead websites, and old versions of any page.

9. https://t.co/h9HNKQiOfG
Tells you exactly why your website is slow with a full report card.

10. https://t.co/wa9s0sjsz5
Rewrites your writing to grade 5 reading level. Cuts the AI fluff in seconds.

People pay $50/month for half of these features inside bigger apps.

These cost nothing.

[Jafar Najafov @JafarNajafov]

Best GitHub repos to replace Cursor for free:

1. Aider
https://t.co/w0489iI8bm

2. Cline
https://t.co/so1XtsrEwP

3. Continue
https://t.co/5Li07nmN0v

4. Roo Code
https://t.co/upnmseSUVQ

5. OpenHands
https://t.co/ghSD6WD8bc

6. Plandex
https://t.co/blgcfplJ1Q

7. Goose
https://t.co/Mh4A3q5Pt9

8. Tabby
https://t.co/jeJUqj7j8r

9. SWE Agent
https://t.co/7XUy2QKZE2

[Ritesh Roushan @devXritesh]

System Design Series - Day 30/30

Deployment Strategies That Prevent Disasters

The most dangerous moment in engineering is deploying to production.

Most junior engineers know one strategy: push it and hope.

Here are 5 deployment strategies that separate junior from senior engineers 👇

1. Strategy 1: Rolling Deployment

The most common and safest starting point.

Instead of updating all servers at once, update one server at a time.

- Server 1: Updated and healthy
- Server 2: Updated and healthy
- Server 3: Updated and healthy

If Server 2 fails health check:
Stop deployment.
Only Server 1 and 2 are affected.
Roll back Server 2.
Server 3 stays on the old version.

Zero total downtime.
Built-in safety net.

2. Strategy 2: Blue-Green Deployment

Run two identical production environments.

- Blue: Current live version (serving all traffic)
- Green: New version (deployed but receiving zero traffic)

Steps:
1. Deploy new version to Green
2. Run full test suite on Green
3. Switch traffic from Blue to Green (instant)
4. If problems: Switch back to Blue (instant rollback)

Zero downtime.
Instant rollback.
Gold standard for zero-risk deployments.

Cost: Double the infrastructure (worth it for critical systems).

3. Strategy 3: Canary Deployment

Named after canary birds in coal mines.

Instead of switching all traffic at once:

- Deploy new version
- Send 5% of traffic to new version
- Monitor for 30 minutes
- If stable: Increase to 20% → 50% → 100%

If problems at any stage: Route 100% back to old version.

Only 5% of users are ever affected.

Used by Netflix, Google, and Amazon for every major release.

4. Strategy 4: Feature Flags

Ship code without activating features.

Deploy the new checkout flow to production.
Feature flag is OFF.

- Internal team: ON for employees only
- Beta users: ON for 10% of users
- Full rollout: ON for everyone

Problems during rollout?
Turn the flag OFF.
Instant rollback without redeploying.

Tools: LaunchDarkly, Unleash (open source), or simple database boolean.

Separates deployment from release.

5. Strategy 5: Database Migration Safety

The deployment strategy nobody teaches but everyone needs.

You cannot change database schema and deploy code simultaneously.

Old code running while migration runs = broken queries.

The safe pattern (backwards compatible migrations):

- Week 1: Add new column (nullable, old code ignores it)
- Week 2: Deploy new code (uses new column if present)
- Week 3: Backfill data in new column
- Week 4: Remove old column

Never break existing queries during migration.

This is why experienced engineers fear schema changes more than new features.

Which deployment strategy does your team currently use?

Reply with the number (1-5).

This concludes the System Design Series.

Thank you for following along for 30 days.

#SystemDesign #Deployment #Backend

[Neo Kim @systemdesignone]

If you're just getting started with system design, learn these 15 concepts:

1 How Virtual Machines Work
↳ https://t.co/QokpR78GgQ

2 Modular Monolith Architecture
↳ https://t.co/VVV6v3KGHJ

3 Redis Use Cases
↳ https://t.co/hZ571ruVeA

4 How RPC Works
↳ https://t.co/yeIgcmAxQx

5 How Apache Kafka Works
↳ https://t.co/8rOy9KgCMo

6 How JWT Works
↳ https://t.co/SZXXrlBsWH

9 How Consistent Hashing Works
↳ https://t.co/7d6EipPcKF

10 How Service Discovery Works
↳ https://t.co/BcL3tgxx1u

11 API Versioning - A Deep Dive
↳ https://t.co/OHAtKSUgVN

12 How Idempotent API Works
↳ https://t.co/afe7ACuSYE

13 How Saga Design Pattern Works
↳ https://t.co/2CffTodOHL

14 How DNS Works
↳ https://t.co/H7hcZnws8N

15 API Design Best Practices
↳ https://t.co/I2ejJ0kbYq

What else should make this list?

===

💾 Save this for later & RT to help other software engineers ace API design.
👤 Follow @systemdesignone + turn on notifications.

[(soon) Dr. JB, Psy.D., LMFT @DopaBim]

How I secured my first non-tech job in Germany 🇩🇪 without certifications or a German language test.

A thread 🧵