Chatbots can be real WoW!! The recent evidence is: ChatGPT. Now that they are more human-like with the latest LLMs (Large Language Models). But these LLMs are Pretrained on their own (HUGE) data. Mere mortals don't have any ways ($$, time, expertise) to train own LLMs. Some do have facility to get fine-tuned on custom corpus, but limited. Custom fine tuning of text documents is being provided by many. Embarking on the journey to master the powerful and sought-after paradigm of RAG, along with multi-modal fine-tuning techniques and Knowledge Graphs. The combination of LLM and RAG (with KG) is an IKIGAI - a concept that the world needs, is willing to pay for, and something you are good at and enjoy. This aligns with Naval’s suggestion of cultivating ‘Specific Knowledge’ - a unique skill set that is untrainable and possessed by few.
This repo is a collection of various PoCs (Proof-of-Concepts) to interface custom data using LLMs. Some notes here
- On various modalities, use cases and domains, for ChatBots
- Prep videos, write Medium Posts (GDE/TH), LinkedIn posts, Youtube channel,
- Make a webpage and store them there as a portfolio, Opensource at Github [Contrary thought: Let me LinkedIn page itself be the webpage, no extra url to maintain]
- Learn some end-to-end hosting platform (LangServe? Langchian with Azure AI?)
- Convert above demos to have user input (disclaimers, limited uploads, $$)
- Make intro videos of all the workshops and courses (use GDE video material)
- Show open source course material, always updating
- MidcurveNN: graph to graph transformation learning
- Indic Languages: OCR for sharada scripts, tokenizer for Sanskrit
- Fine-tuning LLMs with own data using LoRA etc
- Retrieval Augmented Generation (RAG) on own data
- WHY?: World needs, ready to pay, I am good at, and I like it making wow chatbots
- Domain:
- on knowledge graphs, more grounding
- tabular financial data, representation and similarity
- midcurveNN Geometric serialisation and retrieval
- active loop idea of fine-tuning your data
- langchain and llamaindex with any new llm
- bharat gpt, bhashini with sanskrit, do prototype on arthashastra principles
- Specific Knowledge - LLMs, Graphs, Sanskrit
- Enterprise: Google Doc AI, Vertex AI, Microsoft Azure Language AI Services
- Open Source: Langchain (Serve/Smith/Graph), HuggingFace, Streamlit for UI
- Local (secure), no over-the-net API/web calls
- Open source, Free via HuggingFace, Contrib possible
- PoC to Prod, end-to-end
- Python!! end-to-end, with Streamlit as UI
- Huge support, community, opportunities
- Coach: write/talk about it via Medium Stories, Webinars, LinkedIn posts (Mvp ++, advocu ++)
- Passive MicroSaaS income, pay per use, Integrations
- {KG + LLM} chatbot Building LLM based Chatbot on Knowledge Graph
- Knowledge as a Service (KaaS) Building Knowledge Graph from Text and serving it as a Service
- LLM models for Indic (especially Sanskrit) languages. Here is collection of similar efforts going on Awesome AI By Bharat (AABB)
- Do you have unfair advantage:
- Network of founders, influences, for further reach
- Audience: folks who want this app and can pay
- Being early
- Start With a Problem or many problems (don’t tell me your ideas)
- Move from Problems to Solutions, easy, debuggable
- Evaluate Your Solutions
- How is Your Solution Different?
- Talk to Potential Customers
- Start Marketing Before Coding
- Build MVP
- Solves any specific need (pain point) and not anything-and-everything,
- Is it for specific people, 1000 true (paying) fans, say $30 or $3 a month
- Is it a daily need?
-
Focused Development: Micro SaaS businesses focus on serving a niche market or a specific customer segment with a highly targeted software solution[^10^]. This allows for focused development and targeted marketing⁹.
-
Cost-Effective: Micro SaaS businesses operate with minimal resources, leveraging cloud infrastructure and automation tools to streamline operations and keep costs low[^10^]. RAG offers an affordable, secure, and explainable alternative to general-purpose LLMs, drastically reducing the likelihood of hallucination⁴.
-
Customized Solutions: RAG allows businesses to achieve customized solutions while maintaining data relevance and optimizing costs⁶. By adopting RAG, companies can use the reasoning capabilities of LLMs, utilizing their existing models to process and generate responses based on new data⁴.
-
Integration with LangChain: LangChain is a framework designed to simplify the creation of applications using LLMs¹. It can dynamically connect different systems, chains, and modules to use data and functions from many sources, like different LLMs¹. This allows businesses to develop language model-powered software applications that can carry out various activities, including code analysis, document analysis, and summarization².
-
Data-Aware and Agentic: LangChain is data-aware and agentic, enabling connections with various data sources for richer, personalized experiences³. This allows for better interoperability across the board, offering various valuable tools that allow businesses to connect to different vendors (including other LLMs) and integrations with a comprehensive collection of open-source components¹.
-
Access to Various LLM Providers: LangChain offers access to LLMs from various providers like OpenAI, Hugging Face, Cohere, AI24labs, among others¹. These models can be accessed through API calls using platform-specific API tokens, allowing developers to leverage their advanced capabilities to build as they see fit¹.
-
Recurring Profits and Low Risk: With their recurring profits, fewer capital needs, low risk, dedicated customers and minimal operating expenses, Micro SaaS has started attracting many entrepreneurs towards them in recent years¹⁴.
-
Stable Recurring Income: Micro-SaaS businesses are usually location-independent and can be a source of stable recurring income once the product has achieved a loyal user base¹¹.
Sources: (1) What is Micro SaaS And How to Create One In 2024. https://bufferapps.com/blog/what-is-micro-saas/. (2) Complete Guide to Micro-Saas: Build a Profitable Business.. https://blog.payproglobal.com/micro-saas-guide. (3) RAG and LLM business process automation: A technical strategy. https://blog.griddynamics.com/retrieval-augmented-generation-llm/. (4) Retrieval Augmented Generation using Azure Machine Learning prompt flow .... https://learn.microsoft.com/en-us/azure/machine-learning/concept-retrieval-augmented-generation?view=azureml-api-2. (5) What is LangChain: How It Enables Businesses to Do More with LLMs. https://www.bluelabellabs.com/blog/what-is-langchain/. (6) LangChain: A New Era of Business Innovation - Medium. https://medium.com/@tvs_next/langchain-a-new-era-of-business-innovation-7207a44382c9. (7) What is LangChain? A Beginners Guide With Examples - Enterprise DNA Blog. https://blog.enterprisedna.co/what-is-langchain-a-beginners-guide-with-examples/. (8) Top 25 Profitable Micro SaaS Business Ideas in 2022 - StartupTalky. https://startuptalky.com/micro-saas-business-ideas/. (9) Building a Micro-SaaS: Best Tools and Platforms In 2022 - Saastitute. https://www.saastitute.com/blog/building-a-micro-saas-best-tools-and-platforms. (10) Improve LLM responses in RAG use cases by interacting with the user. https://aws.amazon.com/blogs/machine-learning/improve-llm-responses-in-rag-use-cases-by-interacting-with-the-user/. (11) An introduction to RAG and simple/ complex RAG - Medium. https://medium.com/enterprise-rag/an-introduction-to-rag-and-simple-complex-rag-9c3aa9bd017b. (12) Concept of RAG (Retreival-Augmented Generation) in LLM. https://blog.devgenius.io/concept-of-rag-retreival-augmented-generation-in-llm-4f878251b4d1. (13) How To Build a Profitable Micro-SaaS Business in 2024 - BufferApps. https://bufferapps.com/blog/how-to-build-a-micro-saas/. (14) Top 10 Micro SaaS Ideas To Build a Profitable Business in 2024. https://controlhippo.com/blog/micro-saas/.
Not looking for Success, but Wonder!!
- तमसो मा ज्योतिर्गमय : From Dark (hidden in text data) to Light (insights)
- Abhinav Kimothi, RAG Expert: LinkedIn, Projects Portfolio, Website, Medium, LinkedIn Articles, LinekdIn Posts, Company
- Pradip Nichite, Freelancing Expert: LinkedIn, Projects Portfolio, Blog, Youtube, LinekdIn Posts, Company
- Sahar Mor: LinkedIn, Blogs
- LangChain How to and guides
- Building the Future with LLMs, LangChain, & Pinecone
- LangChain for Gen AI and LLMs - James Briggs
- Finetuning GPT-3 David Shapiro ~ AI
- Build overpowered AI apps with the OP stack (OpenAI + Pinecone)
- Learn about AI Language Models and Reinforcement Learning Kamalraj M M
- GPT-4 & LangChain Tutorial: How to Chat With A 56-Page PDF Document (w/Pinecone)
- LangChain - Data Independent
- Node Classification on Knowledge Graphs using PyTorch Geometric
- Geometric Deep Learning
- PyTorch Geometric
- Machine and Language Learning Lab IISc
- Semantic Web India Enables organizations generate value from Data using AI, Knowledge Discovery
- Cambridge Semantics
- Kenome Partha Talukdar. Helping enterprises make sense of dark data using cutting-edge Machine Learning, NLP, and Knowledge Graphs.
- Knowledge graphs
- [Geometric Deep Learning)[https://geometricdeeplearning.com/]
- Learning on Graphs Conference
- Group Equivariant Deep Learning (UvA - 2022)
- Retrieval-Augmented Generation for Large Language Models: A Survey