I failed my devops internship assignment , my fault slightly .
So , if you're looking for purely information/data , I don't think you should continue , because this article is purely a jorunal . Now that we have our motive set , let's continue by articulating what our subsections look like .
What happened ?
My approach to solve
While looking for internships , I saw this hiring assignment (https://github.com/Alchemyst-ai/hiring/tree/main/may-2026/devops/quickstart) on twitter . Now i have very slight clue about devops , but this job market has not yet given me a chance , so I was like why not . ( I had made an aws account 4 days ago )
I tried the assignment with the help of AI obviously , now there were a few blockages like , zero investment mindset ( because it was obvious that only luck could get me selected ) and not enough time ( college has mandatory workshop going on ) .
Regardless we turn on our laptop and get down to business . The project was very simple , an iii (iii.dev) project , that basically runs 3 servers . I had no clue was iii was either just like you , so it basically is an orchestrator that runs python and typescript files and lets them talk to each other using websockets .

Now here you can see , its not that big of project and should be rather easy to deploy , right ?
I create VPC and make subnets and security groups , now the thing was that these 3 services should not directly contact the outside internet but should be proxied through a API gateway .
Now how do you approach this ? You create a NAT gateway and then let it talk to the outside world right ? Right ???? NAT gateways are charged per hour on AWS !

Btw NAT is internet way of managing many IPs at once . Many internal private IPs are managed under one single public IP . Network Address Translation is its full form (Yes I googled ) .
So this was the proposed plan , but since NAT is expensive , so plan B , run an EC2 as NAT . But there were several issues with it as well . I did try with claude and was able to send the data to the NAT EC2 instance but it was not accepting . After a lot of going forth and back , I dropped the idea.
Then I came to know about fck-nat , an AMI ( configuration about a system , like an Image (docker wala) for Operating Systems ) . Its whole purpose is to run as a NAT and convert those internal IPs to public IPs .

I was not able to configure that as well properly , but then I at last just directly change the Security Group's rules and Route table's rule to directly point to Internet (0.0.0.0) . Now that it has internet access , I tried to run those models , Engine worked perfectly , Inference worker connected to it properly but then the caller worker , needed VM . I was running a t2.micro : - )) ( because its free for 750hrs/month ) .
I was done at that point , because I had done the above whole thing about 4 times , the VPC creation and all of its components was exhausting , but I did learn a few things . Even though it was not working , I sent the proposal and the result has not yet come , but I already know it .
BTW I did get 1.18 dollar bill even after all this.
Also I learnt about reachablitiy analyzer (https://docs.aws.amazon.com/vpc/latest/reachability/what-is-reachability-analyzer.html) in aws , pretty useful .
If you liked the article please do comment , thanks for reading ❤️ .
3
1
0