---
title: "Aditya Biranje (@AdityaBiranje) | Hashnode"
description: "AI Researcher. Read the latest articles by Aditya Biranje on Hashnode."
url: "https://hashnode.com/@AdityaBiranje"
---

@AdityaBiranje

# Aditya Biranje

@AdityaBiranje · Joined January 2026

AI Researcher

### [What Is a Causal Decision Model \[CDM\]? Reverse-Engineering Jev, in TypeScript](https://adityabiranje210.hashnode.dev/causal-decision-model-jev-typescript?utm_source=hashnode&utm_medium=feed)

Nobody outside TypeSafe knows how Jev works, but people have probed it. Here is what is published, what is observed, and what is guesswork, plus tested TypeScript for calling it safely.

21 min read · [adityabiranje210.hashnode.dev](https://adityabiranje210.hashnode.dev/causal-decision-model-jev-typescript?utm_source=hashnode&utm_medium=feed)

[Oct 21](https://hashnode.com/posts/causal-decision-model-jev-typescript/6ad8aeb3610a72fdc29e587c)

### [I Trained a Digit Classifier Using Only NumPy. Here's Every Equation Behind the 82% Accuracy](https://adityabiranje210.hashnode.dev/i-trained-a-digit-classifier-using-only-numpy-here-s-every-equation-behind-the-82-accuracy?utm_source=hashnode&utm_medium=feed)

model.fit(X, y) trains a neural network in one line. It also hides every equation that makes learning happen. I wanted to know what was actually running underneath that line, so I built a two-layer ne

8 min read · [adityabiranje210.hashnode.dev](https://adityabiranje210.hashnode.dev/i-trained-a-digit-classifier-using-only-numpy-here-s-every-equation-behind-the-82-accuracy?utm_source=hashnode&utm_medium=feed)

[Aug 15](https://hashnode.com/posts/i-trained-a-digit-classifier-using-only-numpy-here-s-every-equation-behind-the-82-accuracy/6a80f82fadff94d4fb710661)

### [Designing the Ingestion Layer of a Production RAG System](https://adityabiranje210.hashnode.dev/designing-the-ingestion-layer-of-a-production-rag-system?utm_source=hashnode&utm_medium=feed)

How a modular parsing, chunking, and embedding pipeline turns PDFs, HTML, DOCX, PPTX, and TXT files into queryable vectors in Qdrant — and where it still breaks.

27 min read · [adityabiranje210.hashnode.dev](https://adityabiranje210.hashnode.dev/designing-the-ingestion-layer-of-a-production-rag-system?utm_source=hashnode&utm_medium=feed)

[3](https://hashnode.com/posts/designing-the-ingestion-layer-of-a-production-rag-system/6a6281ce6d3f3aafddb25818) [Jul 23](https://hashnode.com/posts/designing-the-ingestion-layer-of-a-production-rag-system/6a6281ce6d3f3aafddb25818)