Mighty Sane Developer
Into AI/ML, Quant, Finance, Programming, Maths & Algo Trading.
Building, breaking & learning things along the way.
Feel free to DM
Hi @X ,
I’m 22, an AI/ML Engineer Building at the intersection of AI, LLMs, Agents, Quant & maths and intelligent systems.
Also into psychology, inference, coding & developer tools.
Let’s connect if you’re building, learning, or thinking about something interesting. #connect
Hi @X ,
I’m 22, an AI/ML Engineer Building at the intersection of AI, LLMs, Agents, Quant & maths and intelligent systems.
Also into psychology, inference, coding & developer tools.
Let’s connect if you’re building, learning, or thinking about something interesting. #connect
Fine Tuning vs RAG, which one would you prefer ?
I have a slight bias towards RAG 😅
You can use a domain specific LLM
Update knowledge without retraining
Swap models without rebuilding everything
Easier to debug what the model actually retrieved
Fine tuning is still great when you need to change the model’s behaviour or teach it a specific task by requires alot of computation and training.
What do you reach for first — RAG or fine-tuning? 🤔
Saw a YouTube video from an AI engineer talks talking about the next wave of voice agents.
The interesting shift is from the usual STT → LLM → TTS pipeline to actual speech-to-speech models.
OpenAI and NVIDIA are already moving in this direction, but there are still some hard engineering problems: observability, turn detection, interruptions, latency, and figuring out what actually happened inside the conversation.
The voice itself is getting good.
Now we need to make the systems reliable and configure the latency part.
https://t.co/LXKGoJjf24