Building things that break, fixing them, then building them better.
I'm a full-stack developer who accidentally fell down the ML rabbit hole and never came back up. Currently neck-deep in my thesis project while maintaining a healthy obsession with backend optimization and automation.
The vibe: Write code that actually works, contribute to open-source and occasionally convince AI models to do my bidding.
AI-powered network operations agent that's smarter than your average monitoring tool. Features:
- Hybrid LLM Architecture: Llama 3.3 70B for reasoning + Scout 17B for data crunching
- AutoML Magic: AutoGluon-powered predictions that actually make sense
- Honest AF: System tells you when it can't help instead of hallucinating answers
- Tech Stack: FastAPI, LangChain, ChromaDB, Docker
- Real-world Impact: 60-70% token optimization while maintaining quality
Because network ops shouldn't require a PhD to predict connection failures.
Building a blazing-fast chat backend because JavaScript is nice, but Go makes things go fast.
Making language learning less painful, one flashcard at a time.
Languages I Abuse Daily:
Python Go JavaScript/TypeScript SQL
Frameworks That Tolerate Me:
FastAPI React Node.js LangChain
ML/AI Shenanigans:
AutoGluon Scikit-learn Pandas ChromaDB LLM Orchestration
DevOps (Because Nothing Ever "Just Works"):
Docker Docker Compose Git Linux
Yes, I know there's probably too much Python Notebooks. No, I will not stop.
- How to make LLMs stop hallucinating (ongoing battle)
- Network performance prediction without breaking the token bank
- Why my Docker containers always fail on first build (mysteries of the universe)
Got a cool project? Found a bug in my code? Just want to talk about why tabs > spaces?
- LinkedIn: https://www.linkedin.com/in/jijo-valiyaveettil/
- Email: v.francisjijo@gmail.com
- Open to: Collaborations, code reviews, debates about the best way to handle errors
"First, solve the problem. Then, write the code. Then, realize the problem was different. Repeat." - Every Developer Ever