From BERT Theory to Real-World RAG
My experience building production chatbots with RAG architecture
#nlp#bert#rag
Recent Work Highlights
01
AI Admin Portal & Chatbot
75%+
Accuracy
Production RAG chatbot with 75%+ accuracy using AWS CDK, QuickSight, and Sonnet 4.
#AWS#CDK#AI#Chatbot
02
Text Extraction Admin Portal
Go-Live
Production
Scalable React + TypeScript portal with CI/CD automation and image processing.
#React#TypeScript#CI/CD
03
Redshift Knowledge Base
POC
Delivered
Integrated Redshift as alternative knowledge base source for POC chatbot.
#Redshift#AWS#KnowledgeBase
04
Data Asset Management Portal
Go-Live
Production
Built from scratch with Playwright E2E testing and VAPT compliance.
#React#Playwright#VAPT
05
Chatbot & Speech-to-Speech
2+
Reuses
Chatbot with multi-model selection, reused across subsequent demos.
#Chatbot#AI#Speech
06
Agronomy Admin Portal
3+
Teams
Cross-team delivery with postMessage auth POC and team mentoring.
#React#TypeScript#Auth
Certifications
Latest Posts
Understanding the Architecture
A deep dive into BERT architecture exploring the Transformer encoder and multi-head self-attention mechanism
#nlp#bert#transformer
How to Determine Sentiment Polarity for Annotation
A practical guide to labeling sentiment polarity for ML annotation
#nlp#sentiment-analysis#annotation
How BERT Works and Why It Helps Code-Mixed Text
A whitebox conceptual view of how BERT works, what it learns during pretraining, and why it fits code-mixed text
#nlp#bert#transformer
Why Pretrained Language Models Matter
Exploring why pretrained language models like BERT are practical when labeled data and compute are limited
#nlp#bert#machine-learning