Voxel51

• Explore, tinker with, and push the boundaries of the latest AI and machine learning technologies to showcase what's possible with open-source tools and frameworks

• Build inspiring prototypes, proof-of-concepts, and example applications that demonstrate creative and impactful uses of AI to the developer community

• Contribute meaningfully to the open source ecosystem by creating and maintaining libraries, tools, datasets, tutorials and guides that enable developers to more easily leverage AI capabilities

• Curate, clean, annotate and publish unique multimodal datasets spanning text, images, audio, video, 3D and other emerging data types to facilitate novel machine learning research and applications

• Create engaging technical blog posts, social media content, videos, and presentations that educate and inspire developers and data scientists to build with AI

•  Collaborate openly with our internal teams and the broader AI/ML community to cross-pollinate ideas, share knowledge, and collectively advance the state-of-the-art

• Evangelize our technology and vision by representing the company at hackathons, conferences, meetups, universities, and other developer-focused events

•  Influence the product roadmap by providing input on the developer experience, uncovering opportunities for new features and improvements based on community feedback and your work

• Continuously learn, experiment with, and provide feedback on beta versions of our latest models and capabilities to help shape the future of our platform

Hacker-in-Residence

LinkedIn

1) Introductory computer vision course for LinkedIn Learning

2) Prompt Engineering with LangChain (releasing March 2024)

3) RetrievalAugemnted Generation with LlamaIndex (releasing Q2 2024)

4) Fine-tuning LLMs with Hugging Face TRL (est release Q3 2024)

5) Text Generation with Hugging Face Transformers (est release Q3 2024)

LinkedIn Learning Instructor

The Artists of Data Science

The ONLY personal growth and development podcast for Data Scientists

Host

Deci AI

Developer Relations Manager

Pachyderm Inc.

Senior Developer Advocate

Comet

ML Developer Advocate

Data Science Dream Job

Principal Data Science Educator

Price Industries

Lead Data Scientist

Bold Commerce

Senior Data Scientist

Emergent BioSolutions

Biostatistician

I create content about deep learning | The DevRel and Deep Learning Dude

Hacking on LLMs, curating multimodal datasets, and writing a book about RAG.

Harpreet Sahota 🥑

sam3 is seriously a step change improvement over sam2

the model is seriously good, glad they brought back text prompts. the automatic segmentation isn't as good as i hoped it to be...but it still works quite well. 

i also got it to output image embeddings and that was cool to visualize in the app

learn more here: https://lnkd.in/gzqJzj6G

🤖 Hacker-in-Residence @ Voxel51| 👨🏽‍💻 AI/ML Engineer | 👷🏽‍♀️ Technical Developer Advocate | Learn. Do. Write. Teach. Repeat.

qwen3vl is literally the best model i've used for video usecases. blows my mind

hosting a virtual event tomorrow about document ai

you may or may not find it interesting, but i guarantee you'll learn something

join me here: https://lnkd.in/dFXPimVc

sony ai released a pretty cool dataset called the fairness human centric image benchmark, super high quality labels

instructions for downloading and parsing the dataset into fiftyone format can be found here: https://lnkd.in/g_VwxJx4

physical AI is top of mind as the next frontier for deep learning. but it'll only succeed if you have good datasets

language models scaled with web-scale text. 

vision models got ImageNet. 

now Voxel51 and NVIDIA are trying to do the same thing for physical AI 

autonomous vehicles, humanoid robots, anything that moves around atoms.

building these systems means millions of hours of testing in high-fidelity sims before you can actually deploy anything. 

petabytes of multimodal sensor data from various modalities

 LiDAR, cameras, radar, IMUs. 

but over 50% of simulations end up completely unusable because the input data is bad.

even small calibration or timing errors between sensors cascade through your neural reconstructions. 

you waste weeks and burn through millions in compute costs before you realize what went wrong.

the new FiftyOne Physical AI Workbench. it integrates with NVIDIA Omniverse and NVIDIA Cosmos and basically sits at the start of your sim pipeline to validate, clean up, and prep your multimodal data before it hits neural reconstruction and synthetic data generation.

what it does:

- automatically audits physical AI inputs across 75+ critical checkpoints

- transforms raw sensor streams into structured, searchable datasets

- triggers accurate neural reconstructions and generates synthetic scene variations

this means fewer wasted compute hours. 

iteration loops go from weeks to hours. 

you build safer, more reliable systems on data you can actually trust.

read more about what we've been building: https://lnkd.in/gC6_zn-X

i recently integrated 4 ocr models into fiftyone as remote zoo models these handle text extraction and document parsing. 

all available as remote zoo sources, you can get started with a few lines of code

different approaches for different needs:

1. mineru-2.5

1.2b params, two-stage strategy: global layout on downsampled image, then fine-grained recognition on native-resolution crops. 

handles headers, footers, lists, code blocks. strong on complex math formulas (mixed chinese-english) and tables (rotated, borderless, partial-border).

good for: documents with complex layouts and mathematical content

https://lnkd.in/dSwimdvp

deepseek-ocr

dual-encoder (sam + clip) for "contextual optical compression." 

outputs structured markdown with bounding boxes. has five resolution modes (tiny/small/base/large/gundam). gundam mode is the default - uses multi-view processing (1024×1024 global + 640×640 patches for details). 

supports custom prompts for specific extraction tasks.

good for: complex pdfs and multi-column layouts where you need structured output

https://lnkd.in/dhJn2h-v

olmocr-2

built on qwen2.5-vl, 7b params. outputs markdown with yaml front matter containing metadata (language, rotation, table/diagram detection). 

converts equations to latex, tables to html. labels figures with markdown syntax. reads documents like a human would.

good for: academic papers and technical documents with equations and structured data

https://lnkd.in/dNdTQYKg

kosmos-2.5

microsoft's 1.37b param multimodal model. two modes: ocr (text with bounding boxes) or markdown generation. automatically optimizes hardware usage (bfloat16 for ampere+, float16 for older gpus, float32 for cpu). handles diverse document types including handwritten text.

good for: general-purpose ocr when you need either coordinates or clean markdown

https://lnkd.in/dCTfGfV8


two modes typical across these models: detection (bounding boxes) and extraction (text output)

i also built/revamped the caption viewer plugin for better text visualization in the app: 

https://lnkd.in/dBebWdga

i've also got two events poppin off for document visual ai:

- nov 6 (tomorrow) with a stellar line up of speakers: https://lnkd.in/gkBEY4_r

- a deep dive into document visual ai with just me: 

https://lnkd.in/dFXPimVc

Open article: Document Visual AI with FiftyOne - November 14, 2025 by voxel51.com

Document Visual AI with FiftyOne - November 14, 2025

hype for this new integration, and proud of the team for shipping it at lightning speed. big ups to Daniel and AdonAI for goin beast mode on this one

🚀 Introducing Physical AI Workbench — integrated with NVIDIA Omniverse NuRec and NVIDIA Cosmos to solve the biggest bottleneck in physical AI.  

More than half of all physical AI simulations fail — not from bad models, but from bad input data. A timestamp off by milliseconds or a calibration drift of a few degrees can cascade through your neural reconstructions, leaving you with unusable results after weeks and millions in wasted compute. 

FiftyOne Physical AI Workbench sits at the start of your simulation pipeline and provides turnkey access to NVIDIA technology to ensure every simulation run begins with trustworthy data: 

✅ Audits 75+ critical checkpoints to ensure every reconstruction starts with validated inputs 
🎯 Enriches raw sensor streams into structured, searchable datasets 
🔗 Integrates directly with NVIDIA Omniverse and NVIDIA Cosmos for high-fidelity reconstruction and synthetic data generation. 

The result: an end-to-end workflow that prevents downstream failures — cutting iteration cycles from weeks to hours and eliminating GPU waste on failed reconstructions. We’re excited to see this workflow power autonomous vehicles NVIDIA DRIVE and robotics NVIDIA Robotics development teams move faster with higher accuracy. 

🔗 Read the full technical breakdown: https://lnkd.in/gTy5-TNF

AI product leader at Google, Author, and community builder making complex AI concepts simple for technical and business audiences. Published author of two books on GPT-3 and Neural Search. Lead and manage an active community of 200K+ AI developers across my social media channels and Unwind AI platform. Creator of Awesome LLM Apps, a #1 top-trending GitHub repository with 80k+ stars featuring practical AI Agents and RAG tutorials.

Building a community of 1M+ AI Developers at Unwind AI, an open AI education platform where 200k+ AI developers today learn to build with AI in just 3 minutes a day:
→ Daily AI news that actually matters
→ Code-first tutorials on AI Agents, RAG and LLMs
→ Open-source AI Agent and RAG apps you can use today with full code implementation

Author of books on: 
→ "GPT-3: The Ultimate Guide To Building NLP Products With OpenAI API"
→ "Neural Search - From Prototype to Production with Jina"

Previously:
→ Built Tenstorrent's developer community from scratch & launched first generation of AI hardware products from AI Accelerator cards to AI Desktop Workstations.
→ Built AI systems for Wall Street hedge funds @Arcesium (DE Shaw)
→ Led data initiatives at 256 Network
→ Tech Leaders Fellow @Plaksha

I share daily tips and tutorials on:
→ Building production-ready LLM apps
→ Implementing RAG systems
→ AI Agents and opensource agentic frameworks

🤝 Let's connect if you're:
→ Building with AI and need practical guidance 
→ Looking for ready-to-use LLM implementations
→ Want to stay ahead of AI developments and grow your career in AI

Follow me for daily AI insights and tutorials on AI Agents, RAG and LLMs.

Shubham Saboo

We just ran the biggest AI Agent course ever 🤯

Here's the list of all the resources we released from Day-1 to Day-5.

250+ pages of production AI Agent insights. 100% free.

📌 Day 1: Introduction to AI Agents (50 pages)
Most agent demos fail in production. This covers the 5 levels of AI agents, from basic reasoning to self-evolving systems, plus the security and architecture patterns that actually work at scale.

Link: https://lnkd.in/dQSzVjR7

📌 Day 2: Agent Tools and MCP (50 pages)
MCP servers can add new tools without your approval. Your poetry bot suddenly gains purchasing power overnight. This breaks down tool design, the Model Context Protocol, and the security risks hiding in plain sight.

Link: https://lnkd.in/dUztxWDA

📌 Day 3: Context Engineering and Memory (70 pages)
Agents with zero long-term memory treat every session like meeting you for the first time. Learn how to build memory systems that actively curate context, not just save conversations.

Link: https://lnkd.in/dMqjwrqF

📌 Day 4: Agent Quality (50 pages)
Traditional QA doesn't work for agents. You can't unit test judgment. This covers the 4 pillars of agent quality (effectiveness, efficiency, robustness, safety) and how to actually evaluate the reasoning process.

Link: https://lnkd.in/dQR3_HkG

📌 Day 5: Prompt to Production (40 pages)
Building an agent takes 2 weeks. Trusting it in production takes 6 months. Learn evaluation gates, circuit breakers, and evolution loops that close the gap between demo and deployment.

Link: https://lnkd.in/dVHSeWmE

With insights from experts including Antonio Gulli, Alan Blount, Michael Zimmerman, Vladimir Vuskovic, Michael S., Kanchana Patlolla, Santiago (Sal) Diaz, Madhurranjan Mohaan, Kimberly Milam, Meltem Subasioglu, Turan Bulmus, Wafae Bakkali, PhD, Dr. Sokratis Kartakis, Gabriela Hernández Larios, Ran Li, Elia Secchi, and Huang Xia 🔥

Kudos to Kanchana Patlolla, Anant Nawalgaria and Polong Lin for the amazing organization 🙌

The best part?
250+ pages. Zero fluff. 100% free.

Repost this so your network doesn't miss it ♻️

AI Product Manager @ Google | Open Source Awesome LLM Apps Repo (#1 GitHub with 80k+ stars) | 3x AI Author | Views are my Own

No alternative text description for this image

I just open-sourced 100+ AI Agent & RAG tutorials that others charge $10,000+ for...🤯

100% FREE. 100% Opensource. Zero paywalls.

Here's what 200K+ AI developers are building with these tutorials:

Starter AI Agents (Pick Your First Project):
• AI Blog to Podcast Agent
• AI Travel Agent (Local & Cloud)
• AI Data Analysis Agent
• AI Music Generator Agent
• AI Meme Generator Agent (Browser)
• Mixture of Agents

Advanced AI Agents:
• AI Deep Research Agent
• AI Investment Agent
• AI Movie Production Agent
• AI System Architect Agent
• AI Product Launch Intelligence Agent
• AI Home Renovation Agent with Nano Banana

Multi-Agent Teams:
• AI Legal Agent Team 
• AI Game Design Agent Team
• AI Teaching Agent Team
• AI Recruitment Agent Team
• AI Real Estate Agent Team
• Multimodal Design Agent Team
• AI Competitor Intelligence Agent Team
• Multimodal UI/UX Feedback Agent Team with Nano Banana

Autonomous Game Playing Agents:
• AI Chess Agent vs Agent
• AI 3D Pygame Agent
• AI Tic-Tac-Toe Agent

Voice AI Agents:
• Customer Support Voice Agent
• AI Audio Tour Agent
• Voice RAG Agent 

RAG Systems (Every Type Covered):
• Agentic RAG with Reasoning
• Agentic RAG with Embedding Gemma
• Corrective RAG (CRAG)
• Autonomous RAG
• Contextual AI RAG Agent
• Vision RAG
• RAG with Database Routing
• Gemini Agentic RAG

Complete Framework Crash Courses:
• Google ADK Crash Course
• OpenAI Agents SDK Crash Course

Every tutorial includes:
→ Step-by-step instructions
→ Real-world application
→ Opensource code on GitHub

Why am I giving this away for free?
Because I believe the best AI knowledge shouldn't be locked behind paywalls.

Get started in 2 simple steps:
1. Subscribe to Unwind AI (for free): theunwindai.com
2. Star the repo: https://lnkd.in/dW6b_dEn

New AI Agents and RAG tutorials added every week.

P.S. I create these AI tutorials and opensource them for free. Your 👍 like and ♻️ repost keeps me going.

Don't forget to follow me Shubham Saboo for daily tips and tutorials on LLMs, RAG and AI Agents.

Google just launched Code Wiki to turn any GitHub URL into instant docs  🔥

Code Wiki automatically generates, updates, and maintains documentation for your entire codebase using Gemini agent.

No more outdated docs. No more hunting through files for hours.

Here's what it does:
• Scans your repo after every commit and regenerates docs automatically
• Gemini chat that knows your entire codebase end-to-end
• Click any concept and jump straight to the exact code file
• Auto-generates architecture and sequence diagrams that stay synced
• Works with any public GitHub repository

Reading and understanding existing code is one of the biggest bottlenecks in software development.

Code Wiki bridges the gap between learning about the code and actually exploring it.

New contributors can make their first commit on Day 1. 
Senior devs can understand new libraries in minutes instead of days.

The best part? 
100% free in public preview.

Private repo support coming soon via Gemini CLI extension.

Link to try it out in the comments.

AI Product Manager @ Google | Open Source Awesome LLM Apps Repo (#1 GitHub with 78k+ stars) | 3x AI Author | Views are my Own

Google's repository of production-grade AI agents just went public 🔥

Agent Garden is now available to all developers.

Agent Garden is a curated repository of ready-to-deploy AI agent samples built with the Agent Development Kit (ADK).

It solves one of the biggest pain points in agent development: the upfront design and integration work that slows teams down.

Here's what you get:
→ Pre-built agent samples with full source code on GitHub
→ One-click deployment to Agent Engine with playground UI
→ Integration with BigQuery, Vertex AI Search, and other cloud services
→ Detailed architecture docs and use case guides
→ Firebase Studio integration for quick customization

The workflow is simple:
Explore → Deploy → Test → Customize

Every sample comes with detailed architecture docs, use cases, and fully open-source code on GitHub.

Link to Agent Garden in the comments!

AI Product Manager @ Google | Open Source Awesome LLM Apps Repo (#1 GitHub with 77k+ stars) | 3x AI Author | Views are my Own

I just released this AI Agent Course for free what others charge $2,000+ for 🤯

A complete crash course on Google's Agent Development Kit with Gemini 3.

No paywall.
No email gate.
No get link in DM

Just drop everything and learn:
•  Agent creation & structured outputs
•  Tool integration (MCP, LangChain, custom functions)
•  Memory systems (in-memory + SQLite persistence)
•  Multi-agent patterns nobody teaches
•  Sequential, parallel, and iterative agentic workflows

9 tutorials. Real code. Real working examples.

Why free?
Because the best things I ever learned came from strangers who asked for nothing in return.

This is me paying that forward.

The repo already has 80,000+ stars.
80,000 developers who said "finally, someone just GAVE us the thing."

Go build something incredible.

🌟 Open Source GitHub Repository: https://lnkd.in/dW6b_dEn

♻️Repost to share it with your network. And follow me Shubham Saboo for more such tutorials.

I just deployed a production-grade AI agent in just 60 seconds 🤯

One command. 60 seconds. That's it

Using Google's Agent Starter Pack created by Elia Secchi. 

Building AI agents is the easy part.

But what follows is a nightmare:
• Setting up infrastructure  
• Configuring CI/CD pipelines  
• Adding monitoring & observability  
• Making it Secure and production-ready

Weeks of work. Dozens of tools. Endless configuration files.

But with Agent starter pack you instantly get: 
• Fully functional agent (ReAct, RAG, or multi-agent)  
• Backend + Frontend ready to go  
• Complete deployment infrastructure  
• CI/CD pipeline configured  
• Monitoring & observability built-in  

Even better:
Already have an agent? Run `enhance` to add production-ready deployment to your existing project.

We're moving from "Can we build AI agents?" to "How fast can we ship them?"

This isn't about prototypes anymore.

This is about getting AI agents into production while your competitors are still wrestling with Terraform configs.

Try it out: The whole thing is open source. 

Zero setup required in Firebase Studio or Cloud Shell.

🌟 GitHub repo: https://lnkd.in/d8jwyAxq

♻️ Repost to share it with your network. And follow me Shubham Saboo for more such tutorials.

Google's Advent of Agents 2025 is now live 🔥

Last month, 1.5 MILLION of you joined the biggest AI Agent course ever on Kaggle. 

Now we are back with another interesting way to learn AI Agents. 

25 days. Zero to Production-Ready AI Agents. 100% free.

Here's what happens EVERY DAY until Christmas:

⚡ One new feature
⏱️ Less than 5 minutes to try
💻 Copy-paste code that actually works
📖 Direct links to docs

No theory. No fluff. Just ship.

What you'll master:
• Gemini 3 (Context engineering, Computer Use, Live API)
• Google ADK (Agent Development Kit)
• Vertex AI Agent Engine
• E2E production templates with Agent starter pack

From quickstart to multi-agent orchestration.
From localhost to production deployment.

The Hub is your permanent archive.
Missed Day 3? Go back.
Need that code from Day 12? It's there.
Building your agent on Christmas Eve? We got you.

👉 https://lnkd.in/dsCF8_Fb

To get started, check out the "Introduction to Agents" white paper, released as part of the 5 Days of Agents with Kaggle.

Read it here: https://lnkd.in/dvuZEDBB

Video intro by Abraham Gomez. A huge thanks to the team building this course for the community Elia Secchi, Alan Blount and many more googlers. 

Day 1 starts now. Let's build AI Agents this December. 

♻️ Repost to share this free and interactive course with your network. And follow this space to stay updated for what's to come in the 25 days.

Introducing Gemini 3 Pro, the new standard for multimodal understanding.

It's not just incrementally better, it's a different category of model. It beats Claude Sonnet 4.5, GPT-5.1 and Gemini 2.5 on almost every benchmark.

I've been testing it for 3 weeks. My first reaction was "wait, that shouldn't be possible."

From analyzing video streams in real-time to solving graduate-level physics, Gemini 3 Pro represents a generational leap in reasoning and capability.

Here is what makes Gemini 3 Pro a huge deal:

🚀 Best model in the world for multimodal understanding: It doesn't just read text, it natively understands video, audio, and code with unparalleled clarity.

🧠 SOTA at complex reasoning: It dives deeper into ambiguous problems, untangling complex logic that stumps other models.

🛠️ Insanely good at agentic tool use & vibe-coding: A force multiplier for developers. Whether it's one-shot prototyping or navigating complex codebases, its "vibe-coding" capabilities allow you to build faster than ever.

⚡ Great at zero-shot generation: Exceptional instruction following that nails the output on the first try, without the need for endless prompt engineering.

🏆 Tops the charts: #1 on the LMArena Leaderboard (1501 Elo), Humanity’s Last Exam, and GPQA.

💬 More direct and factual: It cuts through the noise - delivering concise, factual answers with significantly reduced hallucinations and sycophancy.

Link to the blog in the comment!

AI Product Manager @ Google | Open Source Awesome LLM Apps Repo (#1 GitHub with 79k+ stars) | 3x AI Author | Views are my Own

Entire multimodal AI stack in a single Python library 🤯

Pixeltable is an open-source Python library that replaces your entire multimodal AI infrastructure with a single declarative table interface.

No more juggling databases, vector stores, ETL pipelines, and orchestration tools.

Key Features:
• Multimodal data storage: Images, videos, audio, and documents in one table
• Automatic computation: Define transformations once, they run on all new data
• Built-in vector search: Semantic search without a separate vector DB
• Native AI integrations: OpenAI, Anthropic, Hugging Face, CLIP out of the box
• Time travel queries: Access and query any previous version of your data

Real-world use cases:
• RAG pipelines with document chunking and embeddings
• Object detection across image datasets
• Audio transcription with automatic model management
• Multi-agent workflows with tool calling

Everything persists automatically. All results are versioned. The entire workflow is declarative.

The best part? 
It's 100% Open Source.

Link to the repo in the comments!

_____
♻️ Repost to share this with your network. Follow me (Shubham Saboo) for insights and tutorials on AI Agents and RAG.

AI Product Manager @ Google | Open Source Awesome LLM Apps Repo (#1 GitHub with 75k+ stars) | 3x AI Author | Views are my Own

We just released a 50-page guide on Introduction to AI Agents.

This came out of Google's 5-Day AI Agents Intensive on Kaggle, where we wanted to move the conversation beyond agent demos to real, deployable agentic systems.

Most AI agent demos work great until you try to put them in production.

Then reality hits: hallucinations, security risks, unpredictable behavior, and no clear path to scale.

So we documented everything we've learned about building agents that actually work.

The Core Insight 🧠
An agent isn't just an LLM with function calling.
It's a context curator running in a continuous loop:

Think → Act → Observe → Repeat

What you put into that loop determines success or failure.

The 5-Levels of AI Agents 🪜
→ Level 0: Pure reasoning, no tools
→ Level 1: Connected to external tools
→ Level 2: Multi-step strategic planning
→ Level 3: Multi-agent collaboration
→ Level 4: Self-evolving systems

Three core components of AI Agents ☄️
→ Model (The Brain): When to use Pro vs Flash, model routing
→ Tools (The Hands): RAG, code execution, APIs, human-in-the-loop
→ Orchestration (The Nervous System): Memory, planning, reasoning

The Hard Parts We Go Deep On 🤿
→ Security: Agents as a new class of principals
→ Production: From prototype to enterprise-grade systems
→ Self-evolution: How agents learn and improve over time

You're not a "bricklayer" writing explicit code anymore.

You're a "director" who sets the scene, chooses the cast, provides context, and guides performance.

The AI Agent does the rest.

The best part? 
50+ pages. Zero fluff. 100% free.

With expert insights from Antonio Gulli, Alan Blount, Michael Zimmerman and Vladimir Vuskovic 🔥

Link to the free guide in the comments!

Google just dropped 70 pages on Context Engineering and agent memory 🧠

Day 3 of Kaggle AI Agent Intensive is here.

Most AI agents have zero long-term memory.
They forget everything between sessions.
They treat every conversation like meeting you for the first time.

This isn't just annoying. It's architecturally broken.

What This 70-Page Guide Covers:

1️⃣ Context Engineering (The Foundation):
↳ The difference between prompt engineering vs. context engineering
↳ How to dynamically assemble context
↳ Managing context window limits without losing critical information

2️⃣ Sessions (The Workbench):
↳ What sessions actually are: conversation history + working memory
↳ How different frameworks handle sessions (ADK vs others)
↳ Production considerations: latency, security, data retention
↳ Context compaction strategies: truncation, summarization, pruning

3️⃣ Memory (The Filing Cabinet):
↳ Memory vs. RAG: complementary but fundamentally different
↳ The LLM-driven ETL pipeline: extraction → consolidation → retrieval
↳ Multimodal memory: handling images, audio, and non-textual data

The Critical Insight Most Teams Miss:
Memory isn't just "save the conversation."
It's an active curation system.
All happening asynchronously in the background.

The best part? 
70+ pages. Zero marketing fluff. 100% free.

With expert insights from Antonio Gulli and Kimberly Milam🔥

Link to the free guide in the comments!

Your AI agent just gave away $10K in free products 🤯

It's 3 AM. No one's watching. And you have no idea why.

This is the "last mile" problem killing 8 out of 10 AI agent projects.

Google just dropped a 40-page guide just to talk about that.

Here's the brutal truth:
Building an agent takes 2 weeks.
Trusting it in production takes 6 months.

That gap? It's costing companies millions.

The 3 things you're probably missing:
→ Evaluation gates (no agent ships without passing behavioral tests)
→ Circuit breakers (instant kill switch when things go wrong)
→ Evolution loops (every failure becomes tomorrow's test case)

Real talk:
• A customer service agent tricked into giving products away? Guardrail failure.
• A $50K weekend bill with no explanation? Missing monitoring.
• An agent that worked yesterday but broke today? No continuous evaluation.

These aren't hypothetical. They're happening right now as you read this post.

The solution isn't more AI.
It's better Ops.

Observe → Act → Evolve

Observe the system's behavior in real-time. Act to maintain performance and safety. Evolve the agent based on production learnings. Repeat.

This loop is what separates demos from production systems.

The best part?
Google's Agent Starter Pack gives you production-ready templates with built-in CI/CD, evaluation harness, and observability.

40+ pages. Zero fluff. 100% free.

With expert insights from Dr. Sokratis Kartakis, Gabriela Hernández Larios, Ran Li, Elia Secchi, and Huang Xia 🔥

Link to the full guide in the comments.

♻️ Repost this if someone on your team is shipping agents to production.

Just saw an AI Agent browse the web faster than ever at this MIT Tech event.

Built to scale for millions of web agent operations, not just consumer browsing like ChatGPT Atlas or Perplexity Comet.

It separates inference from execution, translates websites into machine-executable code, and runs thousands of nano-agents in parallel.

No screenshots. No supervision. Just autonomous execution.

TinyFish runs 30 million agentic operations monthly for Google and DoorDash.

Not browser automation. Not web scraping. Operational infrastructure for web agents.

Here's what makes it different:

Most web agents operate in user context. One session. One browser. Human speed.

TinyFish built infrastructure for agents at scale:
• Logs into authenticated systems
• Navigates multi-step workflows across thousands of sites
• Runs hundreds of thousands of parallel agent sessions
• Extracts structured data at machine speed

The difference?
Browser agents are productivity tools. One task at a time.

Agent infrastructure enables what wasn't possible before: comprehensive intelligence across thousands of fragmented sources simultaneously.

This is horizontal infrastructure for agentic systems.

Like AWS for compute. Stripe for payments.
TinyFish for web agents.

Developer can sign up for beta access starting today.

Link in the comments.

taplio

The Battle

Harpreet Sahota 🥑

Shubham Saboo

Followers

Posts

Engagements

Likes

Performance ScoreHarpreet Sahota 🥑's LinkedIn Performance

Performance ScoreShubham Saboo's LinkedIn Performance

Ready to grow your LinkedIn presence?

taplio

The Battle

Harpreet Sahota 🥑

Shubham Saboo

Followers

Posts

Engagements

Likes

Performance Score.css-goub6s{font-size:var(--chakra-fontSizes-md);font-weight:500;color:var(--chakra-colors-gray-600);display:block;margin-top:var(--chakra-space-1);}Harpreet Sahota 🥑's LinkedIn Performance

Performance ScoreShubham Saboo's LinkedIn Performance

Ready to grow your LinkedIn presence?

Performance ScoreHarpreet Sahota 🥑's LinkedIn Performance