Victoriously connected my Windsurf agentic IDE to an EC2 instance that has an A10 GPU to fasten the AI development process. I used a the VPN service provider TailScale to connect my Windsruf to the EC2 instance as a remote connection. This allows my AI environment to remain secure, SOC-2 complaint, and accelerates my coding and testing iteration rate X10! (which saves me time and my client $$$). Now, I get to deploy models inside this A10 NVIDIA GPU, test agentic systems and RAG systems in a near-production environment with the power of agentic IDEs 🔥

Hey folks, Wanted to share a neat tool that helps you cost AWS instances to help you cost AWS resources: https://instances.vantage.sh/ It encompasses EC2, RSD, Elastic Beanstalk, and Redshift instances

Hey folks, The Severless 2-Aurora PostgreSQL AWS engine database engine & instance allows you to effectively create a severless instance for hosting your databases based on time intervals of inactivity (e.g. if the instance has not been active for 5 minutes it goes into "Puause" mode, stopping instance utilization costs). It has pgVector extension support to convert the tabular data into embeddings with pgvector 0.7.0, HNSW indexes, Multiple distance functions,Parallel index building Additionally, it has has the feature of sharding (what a name 🤣), which is the database instance equivalent of horizonal group scaling (i.e. adding more instances of the same CPU type in order to cope with the erratic demands from other servers). My client and I had budgeted more than $200 per month of AWS RSD (i.e. database instance) cost, but would decrease the cost to ~$40 per month! This blurs the line between setting up a testing and production environment database since you would essentially would have the ability to scale down or up without having to click a botton. Recommend you all to check out the database if you work with AWS: https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/aurora-serverless-v2.html

Hey folks, Small W here...my client is transitioning towards hosting his AI infrastructure into AWS. In my attempt to always bring value, I recommended him to apply for credits with AWS's Active Credits program, which targates for startups. He applied and got approved $1,000 worth of free credits! Is not as much as Paul's $25,000 Azure credits recommendation, but something is something :) If you are thinking of creating your AI Infra in AWS, please do apply: https://aws.amazon.com/startups/credits?lang=en-US

I'm deploying an LLM in a server that has primitive features, including a feasible networking infrastructure. The owners of the servers were asking my client to split the cost of creating an API networking infrastructure to allow the LLM (hosted in their A100 GPUs) to other servers. The indicated that the estimated cost of developing this API infrastructure was going to be $12-15,000, developed in a time period of ~2.5 weeks with 4-5 developers with 90-95% chance of success. They wanted my client to go halfsies of the total cost, to which he agreed. A couple of week after the meeting, I proposed to him that I might be able to developed the networking infrastructure from their servers to any other server. To me it just seemed costly to spend $15,000 for an API. Having gained his trust, he agreed. So today, I spend about 2.5 hours on setting up a SSH connection from the Linux based environment of the server to my local computer. The attempt was to connect my Windsurf agentic IDE to the server so I didn't have to use the terminal and or the outdated VS code editor innately integrated in their UI. To by-pass the firewall of the server, I simply rellied on VPN service to make a safe encrypted connection between their server and my computer and.....BOOM! Now I can use my Windsurf agentic IDE from my computer to connect to their server in order vibe code (while as before, I could only hard-code, ew!). This now has to be tested on making inferential request to the deployed LLM but I think the current networking infrastructure will work on this. I should make a video 🤔 about this