learnwithparam's profile picture. โšก Helping engineers go from AI demos โ†’ production. Master Agentic AI apps in 6 weeks ๐Ÿš€ https://learnwithparam.com

Param Harrison

@learnwithparam

โšก Helping engineers go from AI demos โ†’ production. Master Agentic AI apps in 6 weeks ๐Ÿš€ https://learnwithparam.com

๐Ÿš€ Google just launched ADK Visual Agent Builder โ€” a no-code web IDE to design & deploy multi-agent AI systems visually! Part of ADK v1.18, this makes AI agent dev as easy as drag, drop & deploy. AI engineering just got visual. ๐Ÿ‘€ ๐Ÿ‘‰ Follow @learnwithparam for more AI updates


What is your go to solution for AI voice Agentic which can work locally on CPU machines? Any suggestions, I have livekit with kokora TTS with some simple 8b models in mind? But wanted to check whether anyone solved this or seen open source to refer from?


For one of the bootcamp demo (week 2), I created this image 2 image demo (image to colrbook using image model with human in the loop to refine). Ofcourse it can be improved further, but that is the challenge for bootcamp students ๐Ÿ˜ learnwithparam.com/ai-engineeringโ€ฆ

learnwithparam's tweet image. For one of the bootcamp demo (week 2), I created this image 2 image demo (image to colrbook using image model with human in the loop to refine). Ofcourse it can be improved further, but that is the challenge for bootcamp students ๐Ÿ˜
learnwithparam.com/ai-engineeringโ€ฆ

Lately, Iโ€™m a big fan of Autogen agents ๐Ÿ”ฅ Docs are amazing, DX smooth, performance blazing fast. My agent uses their Human-in-the-loop. Curious, any Autogen experience I should watch out for before I pull my hair out later? #AI #Autogen #DeveloperExperience

learnwithparam's tweet image. Lately, Iโ€™m a big fan of Autogen agents ๐Ÿ”ฅ Docs are amazing, DX smooth, performance blazing fast.

My agent uses their Human-in-the-loop.

Curious, 
any Autogen experience I should watch out for before I pull my hair out later?

#AI #Autogen #DeveloperExperience

๐Ÿš€ Google just made RAG stupid simple. Meet Gemini File Search โ€” a fully managed retrieval system. No chunking, no vector DB, no glue code. โš™๏ธ Upload files โ†’ Ask โ†’ Get grounded answers (with citations). ๐Ÿ’ธ Query-time embeddings? Free. RAG devs just got a major upgrade.


Hot take: Building a reliable AI agent has more in common with distributed systems than prompt engineering. Agree or disagree?


Every LLM app fails the same way first: latency. Your prototype feels magical. You deploy it, the model takes 11 seconds, your frontend times out at 10. Suddenly, โ€œjust call the modelโ€ isnโ€™t enough. Hereโ€™s what real engineers do when AI meets engineering ๐Ÿ‘‡ #SoftwareEngineering

learnwithparam's tweet image. Every LLM app fails the same way first: latency.

Your prototype feels magical.
You deploy it, the model takes 11 seconds, your frontend times out at 10.
Suddenly, โ€œjust call the modelโ€ isnโ€™t enough.

Hereโ€™s what real engineers do when AI meets engineering ๐Ÿ‘‡
#SoftwareEngineering
learnwithparam's tweet image. Every LLM app fails the same way first: latency.

Your prototype feels magical.
You deploy it, the model takes 11 seconds, your frontend times out at 10.
Suddenly, โ€œjust call the modelโ€ isnโ€™t enough.

Hereโ€™s what real engineers do when AI meets engineering ๐Ÿ‘‡
#SoftwareEngineering
learnwithparam's tweet image. Every LLM app fails the same way first: latency.

Your prototype feels magical.
You deploy it, the model takes 11 seconds, your frontend times out at 10.
Suddenly, โ€œjust call the modelโ€ isnโ€™t enough.

Hereโ€™s what real engineers do when AI meets engineering ๐Ÿ‘‡
#SoftwareEngineering
learnwithparam's tweet image. Every LLM app fails the same way first: latency.

Your prototype feels magical.
You deploy it, the model takes 11 seconds, your frontend times out at 10.
Suddenly, โ€œjust call the modelโ€ isnโ€™t enough.

Hereโ€™s what real engineers do when AI meets engineering ๐Ÿ‘‡
#SoftwareEngineering

Loading...

Something went wrong.


Something went wrong.