Welcome to the Research & Prototype Repository by @pratikjadhav2726 — a curated and expanding archive of experimental ideas, novel ML techniques, and prototypes.
This repository serves as a launchpad for fresh research directions and modular prototypes that you can fork, remix, and scale. Each folder represents a unique thread of innovation across multiple domains, from RAG to graph ML and more.
| Folder / Notebook | Area | Description |
|---|---|---|
| Agentic Hierarchical RAG | Multi-Agent RAG/NLP | Advanced Retrieval-Augmented Generation with hierarchical, agent-based query decomposition, orchestration, and reasoning. Learn more, Agentic RAG blog, Awesome RAG |
| EHR Multimodal RAG | Healthcare AI / Multimodal | State-of-the-art RAG system for Electronic Health Records, supporting structured, text, and image modalities with clinical reasoning and privacy-first design. See EHR_Multimodal_RAG/README.md, Clinical RAG paper, MedKLIP |
Activation_functions_learnable_novel |
Deep Learning | Novel, learnable activation functions for enhanced expressivity |
KnowledgeGraph |
Graph ML / LLM Tooling | Experiments involving structured knowledge for LLM reasoning |
Model_distillation |
Compression | Survey + implementations of distillation methods for compact model design |
RAG |
Retrieval-Augmented Generation | Building blocks for knowledge-augmented generative systems |
Sector-wise-and-LRP-guided-main |
Explainable AI | Layer-wise Relevance Propagation (LRP) guided model inspection |
Tiered Document Ranking Pipeline for RAG |
NLP / IR | Custom pipeline for multi-stage document reranking in RAG systems |
Voice_Clone |
Speech AI | Spectrogram comparison code for voice cloning systems |
LLM_Finetuning.ipynb |
LLM Ops | Colab notebook for instruction tuning on LLMs |
PEFT_Bloom.ipynb |
PEFT | Parameter-efficient fine-tuning (LoRA, etc.) on Bloom LLM |
Guanaco_Llama_Dataset.ipynb |
Data | Exploratory code using the Guanaco-LLaMA dataset |
Modern research moves fast — and often ideas are lost in the noise. This repo is designed to:
- 💡 Capture sparks of innovation before they’re lost
- ⚙️ Prototype novel approaches in a reproducible way
- 🔗 Connect disparate concepts like graphs, LLMs, and distillation
- 🤝 Foster collaboration and open research culture
This repo is ideal for:
- 🧪 Researchers looking for new baselines or inspiration
- 🧠 Students exploring research topics for papers/theses
- 🛠️ Engineers prototyping ML/LLM tools with real impact
- 🗺️ Practitioners building cross-domain solutions (Graph + LLM, Audio + Diffusion, etc.)
#MachineLearning #LLM #RAG #KnowledgeGraphs #Distillation #VoiceCloning #ExplainableAI #Neo4j #LangChain #ResearchPrototypes #PEFT #LoRA #Bloom #Guanaco #AIResearch #HealthcareAI #MultimodalRAG #AgenticRAG
Got an idea or want to expand one of the threads?
Feel free to fork, open an issue, or submit a pull request!
Released under the MIT License.
“A prototype a day keeps stagnation away.”
This repo is where unfinished brilliance lives until it's ready to shine.