Subham

| Building Agents and Agent's Backend | Provisioning Infra for ML Workloads

[Subham @SubhamHasNoLife]

@mitchellh Where are you moving your codebases and pipelines into? GitLab?

[Subham @SubhamHasNoLife]

@satyanadella @jeffhollan Per Agent SandBox really gives Foundry an edge over what AWS Bedrock agent core is capable of. Will give it a try. However the model access (by number) is less in foundry. With OpenAI going away I am a bit skeptical of Foundry's future directions.

[Subham @SubhamHasNoLife]

@samhogan Really, ReAct, Eval, RAG disappeared!!
Instead of making this post it would be better if you share actual big tech application who are using the alternates.

[Subham @SubhamHasNoLife]

@benln Any plan for Hyderabad?

[Subham @SubhamHasNoLife]

@bryancsk This is more like a forward looking correction I would say. If you take a look at the revenue numbers Figma has gone 41% YOY. Atlassian more than 20%.

[Subham @SubhamHasNoLife]

https://t.co/GEJlxgr0Uw

[Subham @SubhamHasNoLife]

YouTube video on - how do we do Zero downtime migration of On premise File System to AWS. Video is in the first comment

[Subham @SubhamHasNoLife]

@heyblake Not a project but I have a stellar play list for AI engineers - where I have put every ounce of last 4 years experience with AI engineering

https://t.co/gqdEagzEcp

[Subham @SubhamHasNoLife]

New Videos for "DevSecOps from first principles" Series have been dropped. This time we have covered details around Linux. The cloud we generally refer to it is mostly the Linux machines are running on some providers managed platform.

https://t.co/f116N2nbTv

[Subham @SubhamHasNoLife]

@twtayaan Well DevOps starts from the basics of Linux, Shell, Networking, GitOps and understanding of basic build process no Engineer should touch a production cluster without understanding aforementioned concepts.

[Subham @SubhamHasNoLife]

@_vmlops Though people are going gaga on this course - but we need to understand it is mostly for ML researchers
Gaps CS336 Won't Fill
Production Engineering: CS336 doesn't cover deployment, Docker, k8, or cloud
No RAG, agents, LangGraph, or MCP
no Cost Optimization or observability.

[Subham @SubhamHasNoLife]

@NabilMinhaz The very AI Engeneering path is covered in depth
https://t.co/FLBgcQXGau

[Subham @SubhamHasNoLife]

@ashoKumar89 Great post, one thing I would like to add - Seniors always think on failure modes. If you ask any senior/staff engineer - what are the breaking points of the application, even for a very well engineered system they will atleast be able to mention 10 cases.

[Subham @SubhamHasNoLife]

@Python_Dv In enterprise we have used Hybrid approach however number of nuances and issues are there HNSW, IVF, Graph, Vector all the details are covered here
https://t.co/YA3ACOH3JU

[Subham @SubhamHasNoLife]

@akshay_pachaar Thanks for pointing. My bad I scrolled past it.

[Subham @SubhamHasNoLife]

@vitrupo Sometimes I really feel he doesn't know what he is talking in the first place. Software is built on deep domain expertise, do you think curren Amazon or Netflix is elastic. Developed at the moment..it build up in spiral over time.

[Subham @SubhamHasNoLife]

@BharukaShraddha There are few additional consideration as well while take RAG in prod - https://t.co/3WC6ZfLRDX

Would help anyone to have idea around final production checklist for AI Systems

[Subham @SubhamHasNoLife]

@systemdesignone Try this playlist as well - https://t.co/uhVIpajk1D

[Subham @SubhamHasNoLife]

@pvergadia Spot on. Since last year I got reached by multiple of my friend who are either PM or Designer and they have created awesome apps which would solve real world problem.

Unfortunately no one made it to production or any MRR.

Understanding dev eco system and infra is must