<<<<<<< HEAD
This project is a rule-based medical coding assistant that extracts ICD-10 diagnosis codes and CPT procedure codes from clinical notes.
It simulates basic clinical reasoning by identifying multiple conditions and procedures from a single note and returning structured coding outputs.
- Multi-code detection (ICD-10 + CPT)
- Extracts multiple diagnosis codes from a single note
- Detects CPT codes for:
- Evaluation & Management (E/M)
- Imaging procedures (e.g., CT scans)
- Provides explanation ("reason") for each code
- Simple context-aware logic (not just keyword matching)
- Streamlit-based user interface
Input: Patient is a new patient with dental caries.
Output:
- ICD-10: K02.9 — Dental caries
- CPT: 99203 — New patient visit
Input: Patient is a new patient with chest pain and sinusitis. Evaluation and CT scan performed.
Output:
ICD-10 Codes:
- R07.9 — Chest pain, unspecified
- J01.90 — Acute sinusitis
CPT Codes:
- 99203 — New patient office visit
- 71250 — CT chest without contrast
- E/M levels (99202–99205) are estimated, not fully calculated
- Imaging codes may rely on limited context (body part not always explicitly validated)
- Not fully compliant with official ICD-10 / CPT coding guidelines
- Uses rule-based logic (no NLP/ML yet)
- Python
- Streamlit
project_2_medical_coding/ │ ├── app.py ├── data/ ├── services/ └── utils/
pip install -r requirements.txt
streamlit run app.py
📈 Project Status
## Phase 1.5 Complete — Multi-code detection implemented
🔧 Actively improving toward more accurate coding logic
🔄 Next Steps (Phase 2)
Improve E/M level selection using proper criteria
Add coding guidelines (ICD-10 & CPT rules)
Reduce assumption-based coding
Introduce NLP for better clinical understanding
Enhance procedure detection (beyond basic keywords)
💡 Goal
To evolve this project into a more realistic medical coding assistant that mimics real-world coding workflows and decision-making.
=======
# 🎓 AI Portfolio — Aiyesha Talat
AI Engineer focused on building production-ready Generative AI and RAG systems.
Welcome to my AI Engineering Portfolio.
This repository highlights practical AI/ML and Generative AI projects I’ve built while transitioning into an AI Engineering career.
---
## 🔥 Areas of Focus
* Generative AI — LLMs, Prompt Engineering, Agent Orchestration
* Machine Learning — NLP & Computer Vision
* RAG Systems — Retrieval + Knowledge Grounding
* AI on Cloud — Azure OpenAI, Azure AI Search, Cognitive Services
* Responsible AI — Evaluation, Groundedness, Safety
---
## 📌 Portfolio Roadmap
### 🧩 Phase 1 — AI Foundations & Practical ML + GenAI Tools
| Project | Description | Status |
| --------------------------------------------- | ------------------------------------------------------------------------------------------ | -------------- |
| Outlier Detector + AI Slide Generator | Detects outliers in student datasets and auto-generates instructional slides | ✅ Completed |
| 🏥 AI Medical Coding Assistant (ICD-10 + CPT) | Extracts ICD-10 and CPT codes from clinical notes using structured logic with explanations | 🚧 In Progress |
---
### 🤖 Phase 2 — Enterprise AI Apps (RAG, Agents, Semantic Kernel)
| Project | Description | Status |
| ----------------------------------------------------------- | ------------------------------------------------------------------------------------ | -------------- |
| AI Teacher Assistant | Document-aware Q&A assistant using lightweight RAG + Semantic Kernel | ✅ Completed |
| Plant Identification Agent (PlantPal — Offline Local Model) | Streamlit MVP with local PyTorch model for plant image classification + JSON care DB | ✅ MVP Complete |
---
## 📸 AI Teacher Assistant — Demo
---
## 🌿 PlantPal — Demo
### Current Version
* Streamlit UI
* Image-based prediction using local offline PyTorch model
* Confidence score display
* JSON-based care instruction lookup
* Graceful fallback for unknown plant predictions
**Note:** Now using a local offline model for reliable predictions — works without API tokens.
---
## 🏥 AI Medical Coding Assistant — Overview
**Phase:** Phase 1 (In Progress)
**Description:**
An AI-powered application that extracts ICD-10 and CPT codes from clinical notes and provides reasoning for each code.
**Current Capabilities:**
* Clinical note → structured coding output
* ICD-10 code detection
* CPT code detection
* Explanation for each code
**Planned Improvements:**
* Multi-code detection (in progress)
* NLP-based intelligent coding
* Enhanced keyword matching
---
## 🛠 Tech Stack
Python | PyTorch | Streamlit | JSON-based knowledge DB
---
## 🎯 Evaluation Focus
* Groundedness
* Retrieval Relevance
* Accuracy
* Responsible AI Practices
---
## 📂 Repository Structure
```text
ai-portfolio/
├── phase_1/
│ ├── project_1_outlier_slide_gen/
│ └── project_2_medical_coding/
└── phase_2/
├── project_2_ai_teacher_assistant/
└── project_3_plant_identification/- Upgrade Medical Coding Assistant to multi-code + NLP-based system
- Deploy live demos
- Add multimodal workflows
- Expand agent-based systems
LinkedIn: https://www.linkedin.com/in/aiyesha-ai-engineer/ Email: aiyeshat@yahoo.com
c6721ceee99955b22ca7dd9455fa61d43f723a4e