Chasing dreams in code and coffee ☕ https://teapawt.net/t.co/L7joY8RfSV
Joined July 2024
- Tweets290
- Following274
- Followers85
- Likes375
Pinned Tweet
Built a RAG assistant that does it for you - ask a question, get a cited answer pulled from your own PDF library. This is a hybrid retrieval pipeline for scientific papers:
> Dense: cosine similarity over OpenAI embeddings in pgvector → top 20
> Sparse: keyword ILIKE → top 20
> RRF fusion merges both rankings
> Similarity threshold filter (with fallback)
> LLM reranker scores each chunk 0–10
> Generator answers with inline citations -or says 'unsupported' if evidence is weak
Live demo: papyrus.anidoes.dev
GitHub: github.com/anidoesdev/scient…
Somewhere between learning something new, building things, and applying for jobs. ☕💻
Just trying to get a little closer to where I want to be.
GitHub: github.com/anidoesdev
Check out the GitHub repo:
github.com/anidoesdev/MemNes…
And Docs:
memnest.anidoes.dev
AI forgets between chats, and mixes old facts with new ones.
I built MemNest 🪺: memory for AI agents that stays up to date, keeps its history, and lets you see and fix what it remembers.
Works with Claude via MCP. Demo 👇
youtu.be/wX1XcWigqes?si=BvjQ…
How I Distilled GPT-4o-mini Into a 1.5B Model on a 4 GB Laptop GPU
medium.com/@anikajain1307/ho…
From MERN projects → coding agents → AI.
I turned my learning journey into a website.
anidoes.dev
Multi-model approaches in AI break down complex problems better. It's like solving a puzzle with diverse pieces, each bringing unique data power.
VAE model day:
- 1D CNN compresses 30-cycle window to latent vector (dim=16)
- Decoder's MSE shows anomaly; trained on healthy data, degraded struggle
- Delta features = major breakthrough
- Precision=0.38, Recall=0.88, F1=0.53, AUROC=0.75
- A unique trade-off: broad net with some false alarms
Everyone's chasing the single "perfect" model, but what's wrong with embracing the chaos of multi-model solutions?
Learned the hard way: when your AI model acts out, it's a reminder that sometimes, you just have to let it rebel and learn from the chaos.
When you are not able to concentrate but you have to start working on your next project:
A real time anomaly detection system for industrial sensor streams.
Multimodal:
- Numerical time series (vibrations, temperature, pressure, current)
- Audio (mel-spectograms from machine sounds)
- text logs
Reading 20 papers to answer one research question is brutal.
Built a RAG assistant that does it for you - ask a question, get a cited answer pulled from your own PDF library. This is a hybrid retrieval pipeline for scientific papers:
> Dense: cosine similarity over OpenAI embeddings in pgvector → top 20
> Sparse: keyword ILIKE → top 20
> RRF fusion merges both rankings
> Similarity threshold filter (with fallback)
> LLM reranker scores each chunk 0–10
> Generator answers with inline citations -or says 'unsupported' if evidence is weak
FastAPI + Next.js + Postgres + Redis. Fully open source.
github.com/anidoesdev/scient…
