Joined July 2025
- Tweets 185
- Following 109
- Followers 5
- Likes 1,485
10 Photos and videos
May 23
2
58
11 Oct 2025
Hackathon rules only HTML& CSS, i applied JavaScript inside HTML😂 #CoderArmy #RohitNegi #AdityaTandon #Hackathon
2
68
Parvez Khan retweeted
30 Jul 2025
Roti, kapda aur pahad, what more could a man desire?
86
36
614
23,689
30 Jul 2025
Famous words in tech world these days, RAG? and Langchain? Here in simple and short words #RohiNegi #CoderArmy #RAG #VectorEmbedding #Langchain #LLM #techworld
2
1
53
30 Jul 2025
Langchain is an OS framework that simplifies the process of RAG.
-Provides methods for loading and chunking of data.
-Supports multiple embedding models for vector generation.
-Provides methods for easy storing and fetching data from vector DBs.
1
36
30 Jul 2025
RAG is an AI technique to give LLM a specific context according to query.
It converts query into vector and searches relevant documents in vector DBs(like pinecone,chroma).
Then the document is given as context alongside query for desired result.
1
32
29 Jul 2025
If u came from c or java background, js seems weird first but overtime u will settle down to this craziness and see beauty in it's bizzare behavior. #JavaScript #programminglanguages #webdeveloper
1
23
24 Jul 2025
AI knows everything,
but feels nothing.
A baby knows nothing,
but feels everything.
Will we ever create AGI
by making LLMs more and more intelligent?
#AI #AGI #Consciousness #LLM #Spirituality #RohitNegi #ElonMusk #OpenAI #Google
1
17
23 Jul 2025
Here, is the short and clean description of the whole lecture , Vector? IVF? Decision tree method?HNSW?Product Quantization? check out #RohitNegi #AdityaTandon #CoderArmy #VectorSearch #MachineLearningSEO #aisearch
9
2
48
23 Jul 2025
I think even if we calculate exact close nodes with super or quantum computers fast, this approximate nearest neighbours search still will give better results, because as a human being we like little unpredictability, it's interesting to find something new our search feed.
1
27
23 Jul 2025
IVFPQ->>We first apply IVF, find k nearest centroid then apply product quantization. We don't have to calculate distance from whole set just in k cluster's nodes.
1
26
23 Jul 2025
Product Quantization->>We divide the vector in k number of chunks, each chunk is replaced by single number. This number represents 'which' centroid. This way we reduce the memory usage of a vector(ofcourse some info lost),so the calculating distance become optimized.
1
24
23 Jul 2025
When searching, we start from top layer, find closest one of closest one of above layer. We start from randomly chosen close node at start and as we going down the layers , we are coming closer and closer to our desired nodes.
1
21
23 Jul 2025
Hierarchical Navigable Small World(HNSW)->>We create layers of nodes, each layer is treelike structure where every node is connected to it's nearest neighbours, bottom most layer 0 contains all nodes,nodes in every above layer is chosen randomly, no. of nodes are roughly sqrt.
1
21
23 Jul 2025
Decision Tree Method or Binary Space Partitioning->>We divide our search in two parts at median point for each vector entity in cyclic order.
Problem-We shift our focus on one side of node but there is possiblity that the other side node is closer which increases complexity
1
18
23 Jul 2025
IVF->>If we imagine our vector in k dimensional space, there will be many clusters formed. We find(approx) the centers of clusters.We will search few closest cluster and then search for nearest vector among them.
1
18
23 Jul 2025
Vector->>Represents quality of object, each dimension represent a quality.
Relation between two vectors->>
If data is about pattern and preference - Cosine similarity.
If data is about quantity and actual values - Euclidean Distance.
1
18
21 Jul 2025
I studied 1NF, 2NF, 3NF, BCNF in DBMS, they are the most interesting and complex topic if one try to understand and go deep into the definition, but if we see the table with common sense, it seems too easy to normalize it #DBMS #SQL #Normalization
1
11