A few days ago RexiO went live on Product Hunt. What most people saw was a clean launch page and a working product. What they didn't see is where this actually started.
I'm Sijan, the solo founder behind RexiO. I didn't come from a tech family, a CS degree, or any kind of funding. My first real "internet" experience was on a Nokia button phone I could barely afford to put credit on I didn't even understand what the word meant at the time.
Years later, my first paid coding job was a website I built with 20 of mobile balance and zero real knowledge of what I was doing. RexiO itself started as an accident I was building an unrelated social app, added a small AI chatbot just to answer support questions, and that chatbot slowly became more interesting to me than the actual product. I eventually scrapped the whole thing to chase it.
Since then it's been a solo build: writing my own orchestration layer, an intent classifier, 30+ tools, model routing across providers, and fine-tuning our own Bangla-first models on borrowed Colab GPUs plus a full frontend rewrite along the way, more 12-hour bug hunts than I can count, and a relationship ending in the middle of it that honestly redirected the whole direction of my life.
Does RexiO work fully offline for Bangla queries, or does it need a constant internet connection to SpritEX servers to function?
@enes8lsg Great question! RexiO needs an internet connection to work — it's a cloud-based AI assistant, so responses (including Bangla/Banglish understanding) are processed on our servers in real time. Full offline mode isn't something we're planning right now, but we're focused on making the online experience as fast and reliable as possible.
Does it actually run on local Bangla data, or is it just a fine-tuned wrapper around a global model? Curious how much of the underlying intelligence is genuinely homegrown.
@ahmet4316897505 Fair question, and I'll be transparent: RexiO uses a multi-model routing approach under the hood for general capabilities. On top of that, we're actively fine-tuning a large model specifically on Bangla data to improve native language understanding — this is still experimental and evolving right now, but it's a core part of our roadmap. The "local" feel isn't just one model though — it's everything we've built around it: dedicated Bangla/Banglish intent handling, local context retrieval, and cultural tone tuning, which is where most global products fall short.
How does RexiO handle the mix of Bangla and English in the same chat, especially with local slang or code-mixed sentences?
@keremmxlu This was actually one of the hardest problems to get right! RexiO has dedicated language processing tuned specifically for code-mixed Bangla-English sentences — it doesn't try to force everything into one language first, it understands the mix naturally, including common local slang. Glad it's working well for you — always happy to hear specific examples if something trips it up!
Wrote down the full, unfiltered story of how RexiO actually came together — from a Nokia phone in a village to launch day. No polished corporate version, just what actually happened:
From a Nokia Phone to RexiO: The Story Behind Everything
RexiO Is Now Open For Everyone — No Application, No Waitlist
Coming Soon: Image Generation in RexiO
RexiO is getting Image Generation — describe what you want in Bangla or English, right inside your conversation. Free for every user, launching soon.