Hola, Iโm Vedas ๐พ Final-year CSE student at Sathyabama, Chennai. I build backend, cloud and GenAI systems in Java and Python.
Freelancing on cloud and backend since March.
[ here's a small thread on what I do, what I know and what Iโve built ]
[ 1/5 ]
Cloud security scanner in boto3: 5 S3 and IAM findings to 0, then it alerts when I break the bucket again.
No AWS account, it runs on a local emulator (Floci).
[ 1/3 ]
How it works: one Python scanner, 7 rules
Each finding carries its own fix, so --fix and detection share logic
Scanner writes SECFINDING_HIGH log lines -> metric filter -> alarm -> SNS -> SQS inbox
terraform rebuild scans at 0 findings
[ 2/3 ]
Scripts, screenshots and the known gaps (IAM not enforced, Deploy stage can pass without hooks):
https://t.co/x7LWa15NGz
Star it, run it, or send a PR.
[ 3/3 ]
#buildinpublic#devcommunity#opensource
I ran AWS CodePipeline > CodeBuild > CodeDeploy with no AWS account.
Maven build: 17.269 s, page live on localhost:30000.
CodeDeploy hooks didn't run, so I wrote the SSM agent.
[ 1/3 ]
How:
Floci (local AWS emulator) + PowerShell + AWS CLI. "EC2" is a Docker container running Tomcat 9.
Floci queues hook commands for an SSM agent that doesn't exist, so mine polls every 100 ms and runs them via docker exec.
[ 2/3 ]
Still early:
โข Successful LLM path hasn't run against a funded OpenAI account yet
โข No real throughput numbers
Code + browser demo: https://t.co/Y6Vqh83LV2
Stars & PRs welcome.
[ 4/4 ]
#LLMOptimization#SpringBoot#Java#Kafka#OpenSource#BuildInPublic
I built a log-triage pipeline where the LLM is the last resort, not the first step.
Kafka in, WebSocket dashboard out. Four gates, each cheaper than the next: sliding window, log reducer, exact + pgvector cache, then the LLM only on a miss.
[ 1/4 ]
Gate 3 checks cache before calling OpenAI:
Exact fingerprint match = 0 tokens.
Semantic match via pgvector (cosine similarity โฅ 0.90).
Failure path design: vector call fails? Skip. DB down? Go to model. Model fails? FALLBACK incident recorded. Never throws.
[ 3/4 ]
Lastly, would love to connect if any of this overlaps with what youโre working on ๐
Always up for talking cloud, backend, AI agents or hackathons ๐พ
[ 5/5 ]
Hola, Iโm Vedas ๐พ Final-year CSE student at Sathyabama, Chennai. I build backend, cloud and GenAI systems in Java and Python.
Freelancing on cloud and backend since March.
[ here's a small thread on what I do, what I know and what Iโve built ]
[ 1/5 ]