Ritesh Totlani

Software Architect writing about AI, RAG, and the technologies shaping how we build intelligent systems

Mannheim, Germany Joined on Aug 12, 2026

Ritesh Totlani

Oct 28

How many invoices does it take to teach a model?

#machinelearning #llm #ai #python

15 min read

Want to connect with Ritesh Totlani?

Create an account to connect with Ritesh Totlani. You can also sign in below to proceed if you already have an account.

Create Account

Already have an account? Sign in

Ritesh Totlani

Oct 7

Fine-Tuning Reduced My LLM Token Usage by 66% — Here's What I Learned

#ai #llm #machinelearning #python

5 min read

Ritesh Totlani

Sep 23

You Cloned the Aggregate. Why Is Your Read-Only Operation Still Changing State?

#architecture #softwaredesign #ddd #programming

7 min read

Ritesh Totlani

Sep 18

German Sentiment Analysis with BERT: nlptown vs oliverguhr on 20 Real Sentences

#nlp #machinelearning #python #ai

8 min read

Ritesh Totlani

Sep 9

Entities Are Not Rows: A Domain Model That Behaves

#ddd #architecture #dotnet #softwaredesign

6 min read

Ritesh Totlani

Aug 27

I Tested 4 RAG Chunking Strategies Everyone Recommends. 2 Were Quietly Broken.

#ai #rag #llm #python

6 min read

Ritesh Totlani

Aug 20

Not All Multilingual Embedding Models Are Equal — Testing 5 of Them

#ai #machinelearning #rag #python

14 min read

Ritesh Totlani

Aug 12

Building a RAG System From Scratch — Four Components, One Working Pipeline

#ai #python #rag #machinelearning

5 min read

DEV Community

Ritesh Totlani

Badges

Writing Debut

How many invoices does it take to teach a model?

Want to connect with Ritesh Totlani?

Fine-Tuning Reduced My LLM Token Usage by 66% — Here's What I Learned

You Cloned the Aggregate. Why Is Your Read-Only Operation Still Changing State?

German Sentiment Analysis with BERT: nlptown vs oliverguhr on 20 Real Sentences

Entities Are Not Rows: A Domain Model That Behaves

I Tested 4 RAG Chunking Strategies Everyone Recommends. 2 Were Quietly Broken.

Not All Multilingual Embedding Models Are Equal — Testing 5 of Them

Building a RAG System From Scratch — Four Components, One Working Pipeline