Commit d26704

2025-10-16 04:35:46 Prashant Kumar: Update main page
Projects/SkillShift.md ..
@@ 11,28 11,28 @@
# 🔎 **Overview**
- **SkillShift** is an **AI-powered shifting skills seamlessly across roles** that automatically captures, organizes, and transfers critical knowledge within an organization.
- 👉 Ensuring that when employees leave, change roles, or new hires join, no expertise is lost and knowledge is instantly accessible.
+ **SkillShift** is an AI-powered organizational knowledge assistant that helps companies capture, organize, and retrieve insights from internal communication platforms like **Google Chat** and **Telegram**.
+ It allows **admins** to manage employees, data sources, and topic-based tags, while **employees** can query information using an AI chatbot powered by **Retrieval-Augmented Generation (RAG)**.
+ Messages are sanitized, embedded, and stored for semantic search, ensuring **accurate and context-aware answers**.
+ With **secure role-based access**, SkillShift enables teams to find relevant knowledge quickly, reducing dependency on manual information transfer.
+ The tool serves as a **smart, centralized knowledge hub** that enhances **transparency, collaboration, and productivity** across the organization.
+
#### ✨ **Key Characteristics**
1. 📝 Always-On Knowledge Capture
2. 🤖 AI-Powered Summarization & Q&A
- 3. 🔗 Knowledge Graph of People, Projects, and Decisions
- 4. 📦 Automated Handover Packages
- 5. 📊 Succession Planning & Risk Insights
- 6. 🚀 Onboarding Accelerator
+ 3. 🔗 Role-Based Access Control (RBAC)
+ 4. 📦 Tag-Based Organization
---
# ❌ **The Problem**
1. Huge knowledge loss when employees resign or switch roles.
- 2. New hires take weeks/months to ramp up due to scattered tribal knowledge.
- 3. Single points of failure: only one person knows a critical system.
- 4. Past decisions buried in Jira/Slack → teams reinvent solutions.
- 5. No visibility for leaders into knowledge distribution.
- 6. Manual documentation is outdated and inconsistent.
- 7. Experts are overloaded answering repetitive questions.
+ 2. Scattered and unstructured information across multiple chat platforms.
+ 3. Single points of failure: only one person knows a critical system.
+ 4. Manual documentation is outdated and inconsistent.
+ 5. Experts are overloaded answering repetitive questions.
---
@@ 41,36 41,33 @@
#### ✅ **What does it do?**
- **For Employees:** Smooth onboarding, faster learning, less frustration.
- **For Teams:** Reduced dependency on single experts, better collaboration.
- - **For Managers:** Clear risk visibility, succession planning, workforce insights.
- - **For Organizations:** Preserve IP, resilience, and smoother transitions.
+ - **For Managers/HR/Admins:** Preserve IP, resilience, and smoother transitions.
#### ⚙️ **How does it work?**
#### 🏗️ Architecture
- - **Data Sources** → Docs, chat tools
- - **Data Processing** → ETL pipelines, skill/role normalization
- - **Knowledge Graph** → employees ↔ skills ↔ roles ↔ successors
- - **AI Layer** →
- - NLP for skill extraction/matching
- - ML for attrition risk & readiness scoring
- - LLM chatbot for Q&A
- - **App Layer** → dashboards, chatbot, alerts (Google Chat/Telegram)
+ - **Frontend (React):** Employee and Admin dashboards with AI chatbot interface.
+ - **Backend (FastAPI):** REST APIs for authentication, data management, and AI query handling.
+ - **Database (PostgreSQL + pgvector):** Stores employees, tags, messages, and embeddings for semantic search.
+ - **NLP Worker:** Cleans and embeds chat messages for AI retrieval.
+ - **Docker Compose:** Orchestrates all services for seamless deployment.
#### 🔄 **Data Flow**
- 1. Collect & clean data from various sources
- 2. Build knowledge base of roles, skills, projects, etc.
- 3. Run AI models → risk prediction & readiness scoring
- 4. Deliver insights via dashboards + chatbot
+ 1. **Ingestion:** Messages are fetched from Google Chat or Telegram data sources.
+ 2. **Sanitization:** Text is cleaned to remove PII, links, and noise.
+ 3. **Embedding:** Cleaned text is converted into vector embeddings using NLP models.
+ 4. **Storage:** Messages and embeddings are saved in PostgreSQL (pgvector).
+ 5. **Retrieval:** When a user queries, embeddings are generated and similar messages are retrieved for AI-generated answers.
---
## 🚀 **What makes it innovative?**
- 1. Captures knowledge beyond HR data → chats, code, unstructured docs.
- 2. Automated handover with **AI-clone / chatbot of past work**.
- 3. Visual knowledge graph showing ownership & linking artifacts.
- 4. Continuous ingestion & updates (vs periodic HR efforts).
- 5. Contextual Q&A over actual work (vs static profiles).
+ 1. **AI-Driven Retrieval:** Uses RAG to deliver context-aware, factual answers.
+ 2. **Tag-Based Access:** Organizes knowledge dynamically by topics or projects.
+ 3. **Automated Knowledge Capture:** Extracts insights directly from internal chats.
+ 4. **Privacy-First Design:** Sanitizes and anonymizes all data before storage.
+ 5. **Seamless Integration:** Connects with existing chat tools like Google Chat and Telegram.
---
@@ 80,7 77,7 @@
- **Framework & Libraries:** LangChain
- **Frontend:** React.js + Tailwind (dashboards, chatbot UI)
- **Backend:** FastAPI (Python) + JWT auth
- - **Database:** PostgreSQL (pgvector) + Elasticsearch
+ - **Database:** PostgreSQL (pgvector)
- **Integrations:** Telegram Bot API, Google Chat API
- **Deployment:** Docker + GitLab CI/CD (EC2/DigitalOcean)
@@ 90,11 87,10 @@
```mermaid
graph LR
- A[📂 Data Sources → Chats, Docs, Code ] --> B[⚙️ Data Ingestion Pipelines → APIs, ETL, LLM extractors]
- B --> C[🗄️ Knowledge Store → Postgres + pgvector + ElasticSearch]
- C --> D[🔗 Knowledge Graph → Employees ↔ Projects]
- D --> E[🤖 AI Services Layer, LLM Q&A - LangChain, Succession Risk, Handover Gen]
- E --> F[💻 Client Interfaces, Web Dashboards - React, Chatbot - Telegram/GC ]
+ A[📂 Data Sources → Chats, Docs, Code ] --> B[⚙️ Data Ingestion Pipelines → Remove PII, links, and noise]
+ B --> C[🗄️ Knowledge Store → Postgres + pgvector]
+ C --> D[🤖 AI Services Layer, LLM Q&A - LangChain]
+ D --> E[💻 Client Interfaces, Web Dashboards - React, Chatbot ]
```
## 🧩 **Challenges & Learnings**
@@ 105,10 101,15 @@
## 📅 **Future Roadmap**
- **Short-term (MVP):**
- - Integration with **Google Chat & Telegram**
- - AI-powered **chatbot for Q&A**
- - **Handover Package Generator**
- - Dashboards: Employee Profiles & Succession Pipeline Dashboard
+ 1. **Employee & Admin Roles:** Secure login with role-based access (RBAC).
+ 2. **Tag Management:** Admins can create, update, and assign tags to employees.
+ 3. **Data Source Integration:** Configure Telegram or Google Chat groups as data sources.
+ 4. **Message Ingestion:** NLP worker cleans and stores messages with embeddings.
+ 5. **AI Chatbot (RAG):** Employees query data and get contextual, AI-generated answers.
+ 6. **Dashboards:**
+ - **Employee Dashboard:** Profile and chatbot view.
+ - **Admin Dashboard:** Manage users, tags, and data sources.
+ 7. **Dockerized Setup:** Complete stack deployable via Docker Compose.
- **Medium-term:** To be added
- **Long-term:** To be added
@@ 130,4 131,4 @@
## 🏷️ **Categories**
- #project #ai-hackathon-2025 #HR #SuccessionPlanner
+ #project #ai-hackathon-2025
0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9