Subham patnaik

[Subham patnaik @Subhammmmmmmmm]

Also, I’m open to ideas

What features would make an AI-powered web scraping project more interesting or useful?

[Subham patnaik @Subhammmmmmmmm]

Day 16/100

Today’s learning:
• Started a new web scraping project
• Set up the development environment
• Configured API keys and project dependencies
• Planned the project architecture and workflow

#BuildInPublic #WebScraping #Python #GenAI #AIEngineering

[Subham patnaik @Subhammmmmmmmm]

Today’s takeaway:
The quality of an LLM’s output often depends more on the prompt than the model itself.
A well-structured prompt can improve accuracy, consistency, and reasoning without changing the underlying model.

Still exploring AI Engineering deeply

[Subham patnaik @Subhammmmmmmmm]

Day 15/100

Today I learned:
• Prompt Engineering basics
• System, User, and Assistant messages
• How prompts influence LLM responses
• Temperature and token parameters

Also:
• Solved 2 LeetCode problems

#BuildInPublic #PromptEngineering #LLM #GenAI #AIEngineering

[Subham patnaik @Subhammmmmmmmm]

@shivi1026 @ShradhaKhapra_ @ApnacollegeX Great consistency, keep crushing it!

[Subham patnaik @Subhammmmmmmmm]

@itsmenikhitha Nike Dunks retro have an aura among college students

[Subham patnaik @Subhammmmmmmmm]

@vadym_petryshyn Good catch!
I’m using RunnableWithMessageHistory mainly for learning the fundamentals of chat memory. I’ll definitely explore the newer LangChain-recommended alternatives as I go.

Thanks for the heads-up!

[Subham patnaik @Subhammmmmmmmm]

Day 14/100

Today I learned:

• How chatbots can remember previous conversations using chat history
• Session-based memory management with RunnableWithMessageHistory
• How prompt templates help guide LLM behavior with predefined instructions
#BuildInPublic #LangChain #LLM

[Subham patnaik @Subhammmmmmmmm]

Today's takeaway:

An LLM doesn't remember anything by default.
Memory is created by storing conversation history and associating it with a session ID.
Combined with prompt templates, this allows us to build chatbots that can maintain context and respond more consistently.

[Subham patnaik @Subhammmmmmmmm]

Today’s takeaway:

LLMs are only as good as the context they receive.

A good retriever can often improve answers more than a larger model because it delivers the right information at the right time.

Open to suggestions and resources on RAG and LLM Engineering

[Subham patnaik @Subhammmmmmmmm]

Day 13/100
Today I learned:
• Semantic Search
• Similarity Search in Vector Databases
• Metadata Filtering
• Retriever concepts in LangChain
Solved:
• 3 LeetCode problems
#BuildInPublic #RAG #LangChain #LLM #GenAI

[Subham patnaik @Subhammmmmmmmm]

Day 1/20 for learning recursion
1. sort a stack using recursion
2.reverse a stack using recursion
3. LC:606:Construct String from Binary Tree

[Subham patnaik @Subhammmmmmmmm]

@RajGupta245462 Damn good point man
Let’s connect

[Subham patnaik @Subhammmmmmmmm]

Day 9/100

Today I learned:
• Data Collection for LLMs
• Data Accumulation Techniques
• The importance of high-quality datasets in AI systems

Also:
• Cleared 3 interview rounds
• Solved 2 LeetCode problems

#BuildInPublic #LLM #GenAI #AIEngineering #LeetCode

[Subham patnaik @Subhammmmmmmmm]

@AayushSainii Thanks G❤️

[Subham patnaik @Subhammmmmmmmm]

Today’s takeaway:

LLMs don’t memorize your documents.

They convert text into embeddings and store them in vector databases like FAISS. When a query arrives, similarity search retrieves the most relevant chunks, which are then sent to the LLM as context.

[Subham patnaik @Subhammmmmmmmm]

Day 12/100 🚀

Today I learned:
• Vector Stores
• FAISS (Facebook AI Similarity Search)
• Similarity Search
• Embedding Storage & Retrieval
• How vector databases help in RAG applications
Lc:509 and 209 solved today

#BuildInPublic #RAG #FAISS #LLM #LangChain #GenAI

[Subham patnaik @Subhammmmmmmmm]

Key takeaway:
Learning how embeddings, LLMs, and LangChain work together gave me a better understanding of building AI-powered applications

[Subham patnaik @Subhammmmmmmmm]

Day 11/100
Today i have learned about:
•Converting the chunks into embeddings
• different llm like Ollama and open api
• went through the langchain documentation
#BuildInPublic #LLM #LangChain #GenAI #AIEngineering

[Subham patnaik @Subhammmmmmmmm]

Today’s takeaway:

Chunking affects how well an LLM retrieves and understands context.
Learning different chunking strategies helped me understand RAG systems better.
Open to tips and resources on RAG and LLM Engineering

[Subham patnaik @Subhammmmmmmmm]

Day 10/100

Today I learned:
• Document chunking for LLMs
• Fixed-size, Recursive, and Token-based chunking
• How recursive splitting helps preserve context

Also:
• Prepared for a Cyber Security interview

#BuildInPublic #LLM #RAG #GenAI #CyberSecurity