Skanda Vivek

Senior Data Scientist in NLP. Recognizing and solving problems in an increasingly connected world. Creator of

[Skanda Vivek @SkandaVivek]

🎉Introducing AnswerChatAI🎉to answer context specific questions in 3 easy steps! https://t.co/Q8OfEED14Q

Step 1: Upload URL or text and hit search.

Step 2: Ask a context specific question and hit query.

Step 3: Get your answer!

Try it out - It's free 😊
#ChatGPT #GPT3 #AI

[Skanda Vivek @SkandaVivek]

@safwaankay would love to try this, thx!

[Skanda Vivek @SkandaVivek]

Learn how to:

🔍 Extract text from PDFs
🧩 Chunk documents for LLM processing
🔑 Implement both keyword-based retrieval (Okapi BM25) and vector embeddings

The best part? We build everything from the ground up! No prior RAG experience needed. 💪

https://t.co/JHxxQFDPrm

[Skanda Vivek @SkandaVivek]

@universal_so @jxmnop Not really- the price offerings seem head to head at least when I checked(both Miro and lucidchart have free options for a very limited number of boards - like 3)

[Skanda Vivek @SkandaVivek]

@gunsnrosesgirl3 Another way of thinking about it- the 3 on top become 2 below- so there’s still 1 left 😊 (the entire bar got shorter by 1 small bar

[Skanda Vivek @SkandaVivek]

🚀 Exciting News in the Vector DB Space! 🚀

Upstash has recently introduced its own flavor of vector DB, offering a delightful balance between ease of usage and visually appealing analytics of performance. 🚀💻

See my blog for more! https://t.co/yyVfkZReXR

[Towards Data Science @TDataScience]

Explore recent research into Self-RAG, a novel approach that boosts the performance of "vanilla" RAG, by following along @SkandaVivek's accessible writeup. https://t.co/RBcmuZj3mx

[Skanda Vivek @SkandaVivek]

@levelsio If 1 person signs up for each then aren’t you on the hook to build 30 products that folks pre-ordered?

[Harini Nagendra @HariniNagendra]

Daughter's art for a project - yes, we love libraries in this house 😊

[Skanda Vivek @SkandaVivek]

@denny_zhou Does this work during coding challenges as well? I find that by half-time I'm faced with the hard decision of - do I just aim to wrap things up in a coherent way, or do I spend some more time doing more exploration🤣

[Skanda Vivek @SkandaVivek]

@mathemagic1an @denny_zhou This parrot needs a buddy 😂

[Center for Governance of Tech & Systems (GoTech) @GoTechUMD]

(How) will generative #AI change education? GoTech Research Scholar @SkandaVivek argues that reasoning is where humans shine & AI fails making it a weakness that educators can exploit to reduce cases of misuse.

Check out the GoTech Insights blog post: https://t.co/QkahdNlUPY

[Skanda Vivek @SkandaVivek]

Models like ChatGPT/GPT-4 work great in many use cases, but there are many reasons companies are exploring open-source LLMs including:

- Data, privacy concerns with closed APIs
- Domain specific use-cases

Read more in my blog!
#LLM #opensouce #NLP

https://t.co/pI7ZaRaO6G

[Skanda Vivek @SkandaVivek]

OpenAI introduced "function calling" that has the potential to completely transform how we write code and interact with machines. 🔄💡💻

Imagine querying databases using unstructured prompts on instead of SQL!

Curious? 🤔💡Check out my blog post!

https://t.co/BojNHL5y0b

[Skanda Vivek @SkandaVivek]

@InternetH0F Metallick

[Skanda Vivek @SkandaVivek]

Small perks of being a dad!

[Towards Data Science @TDataScience]

When Should You Fine-Tune LLMs? by @SkandaVivek https://t.co/kKYd7jakcK

[Towards Data Science @TDataScience]

LLM Economics: ChatGPT vs Open-Source by @SkandaVivek https://t.co/rQ2h7UV3DS

[Skanda Vivek @SkandaVivek]

@rachel_l_woods Stats?😆

[Skanda Vivek @SkandaVivek]

@SahilBloom What does it mean to spend 1/2 hour per day “with children” when you are 15? Are you the child in that case- but aren’t you spending 24 hours with yourself? Or are you referring to a specific age of children?

I’m tempted to draw a conclusion- but that would be deceiving

[Skanda Vivek @SkandaVivek]

@pmddomingos Multimodal- integrate language with vision, audio, physical touch, figure out how to ingest multimodal data and reduce the net data needed - and make it scalable over the next 10 years