Skip to content
View ilaa-1505's full-sized avatar
  • Bengaluru

Block or report ilaa-1505

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
ilaa-1505/README.md

Hi, I'm Ilaa Chenjeri

I'm a recent grad who got into AI because I wanted to understand how LLMs actually work, not just call an API and move on.

Most of my projects start with a question I couldn't find a good answer to. Why does RAG fail silently? Why does the top-ranked result from hybrid search feel wrong once you re-read the original prompt? Why does Spotify know your listening history but not what you were going through when you hit repeat?

That curiosity tends to turn into code.

What I work with

Python · HuggingFace Transformers · LangChain · FAISS · BM25 · FastAPI · React · Three.js

Things I've built

Smart Bollywood Song Recommender You describe your exact moment. One line. It finds the song where a specific lyric proves it understood you, not your mood, not your genre, your moment. Hybrid search, dual LLM pipeline, 3D constellation UI. Live on HuggingFace.

Explainable RAG System RAG that shows its work. Full debug UI so you can see why each chunk was retrieved, how it ranked, and where the answer actually came from. Built because black-box retrieval is a real problem, not a hypothetical one.

LLM Visualizer Peek inside GPT-2. Tokens, attention maps, embedding space, next-token probabilities. All local. Built this to understand what was actually happening inside the model. Turns out it helps others too.

Error Clustering System Takes thousands of noisy logs, clusters them into meaningful failure groups, names the root causes, tells you where to look first. Log noise is a real engineering problem.

AI Debate Arena Two agents argue opposite sides of any topic. A judge scores each round on logic, evidence, and rebuttal. What happens when LLMs have to persuade each other.

Right now

Digging into agent architectures and mechanistic interpretability. Looking for a full-time AI/ML role, onsite or hybrid.

Get in touch

If you're working on something interesting in the LLM space, I'm always up for a conversation.

Pinned Loading

  1. Smart-Bollywood-Songs-Recommendation Smart-Bollywood-Songs-Recommendation Public

    A Bollywood song recommender that understands the exact moment behind your words using hybrid search, dual LLM reasoning, and a cinematic 3D constellation to find the one lyric that truly gets you.

    JavaScript

  2. Explainable-RAG-System Explainable-RAG-System Public

    Explainable RAG system built on HuggingFace Transformers docs. BM25 + vector search + MMR + reranker with full debug UI

    Python 1

  3. LLM-Visualizer LLM-Visualizer Public

    An interactive tool to visualize how LLMs work internally through tokens, embeddings, probabilities, and attention maps using a locally running GPT-2 model.

    Python 1

  4. Error-Clustering-System Error-Clustering-System Public

    A log intelligence system that clusters thousands of noisy logs into meaningful failures, names the root causes, and tells you exactly where to investigate first.

    JavaScript

  5. AI-Debate-Arena AI-Debate-Arena Public

    An AI debate arena where two autonomous agents argue opposite sides, a judge scores every round in real time, and ideas survive only through logic, evidence, and rebuttal.

    JavaScript