# Linkedin Profile
I built AI_ChatBot as an HTML project: an AI-powered chatbot using LangChain, Flask, and the Gemini API for conversational responses and real-time interaction. The README describes the project but does not specify implementation details, deployment steps, API-key configuration, or whether the proposed improvements have been started.
My AI_ChatBot README describes it as Flask AI chatbot using Gemini API with LangChain.
The features listed for my AI_ChatBot are Real-time AI chat responses; LangChain integration; Gemini API support; Flask-based web interface; Docker support for deployment; Hugging Face deployment ready.
The technologies listed for my AI_ChatBot are Python, Flask, LangChain, Gemini API, Docker, Hugging Face.
The project structure lists templates/, app.py, requirements.txt, Dockerfile, and runtime.txt. The README lists Docker deployment support and Hugging Face deployment readiness as capabilities.
I can install and launch AI_ChatBot with these commands, in order: git clone https://github.com/dipanshu-2106/AI_ChatBot.git; cd AI_ChatBot; pip install -r requirements.txt; python app.py.
The future improvements listed for my AI_ChatBot are Chat history memory; Voice assistant integration; PDF document chatbot; Multi-language support.
My AI_ChatBot repository dates are Started 2026-06, last updated 2026-06.
My GitHub profile is @dipanshu-2106. The repositories listed here span data science, AI, dashboards, and interactive applications; this adds profile and repository details to my existing project context.
My GitHub profile says: On GitHub since 2025. 2 followers.
The listed repository languages are Python (5), Jupyter Notebook (4), HTML (1), JavaScript (1).
My SWYNEX-Final-Data-Science-Project repository is described as Jupyter Notebook. Last updated 2026-09.
My SWYNEX-Modeling-and-Evaluation repository is described as Jupyter Notebook. Last updated 2026-09.
My EduGuard-AI repository is described as AI-powered Student Dropout Risk Prediction System. Python. Last updated 2026-09.
My EduGuard-AI README describes a full-stack web application for predicting student dropout risk with machine learning. Its tagline is “Predict. Prevent. Protect Every Student's Future”. The project repository and its existing summary are listed above.
My EduGuard-AI frontend stack is Next.js 15, TypeScript, Tailwind CSS, Shadcn UI, Recharts.
My EduGuard-AI backend stack is Python FastAPI, SQLAlchemy, JWT Auth.
My EduGuard-AI database is PostgreSQL (Neon Cloud).
My EduGuard-AI README lists CatBoost (99.74% accuracy), LSTM, DistilBERT NLP as its ML models.
- My EduGuard-AI README lists these features: AI Risk Prediction (Single + Bulk); Role-based Access (Admin, Teacher, Student, Counselor); Real-time Analytics Dashboard; Achievement Badges for Students; Counseling Session Management; Automated Notifications.
My EduGuard-AI README identifies me as Dipanshu Gupta — B.Tech CSE 2024-2028, AKS University, Satna MP.
My EduGuard-AI README directs readers to SETUP.md for installation instructions. It does not provide the dataset, evaluation procedure, or test conditions behind the accuracy claim, nor does the supplied text explain how LSTM and DistilBERT are used or detail the roles' permissions.
My SWYNEX-Exploratory-Data-Analysis repository is described as Jupyter Notebook. Last updated 2026-09.
My SWYNEX-Data-Preparation repository is described as Jupyter Notebook. Last updated 2026-09.
My DataPilot-AI repository is described as Upload CSV/Excel → Auto Clean → EDA → AI Insights → AutoML → Forecasting → PDF Reports; built with React, FastAPI, Groq LLM. Python. Last updated 2026-09.
My Skin-cancer-detector repository is described as ResNet50 classifies skin lesions as Cancerous or Non-Cancerous from an image; built with TensorFlow/Keras and Flask. JavaScript. Last updated 2026-09.
My Skin Cancer Detection project uses ResNet50 to classify skin lesions as Cancerous or Non-Cancerous from an image.
My Skin Cancer Detection project is intended for educational, research, and preliminary screening purposes only.
Users can upload a skin image or capture an image using their camera and receive an AI-based prediction with confidence scores in my app.
My project presents a Cancer probability score.
My project also presents a Non-Cancer confidence score.
My Skin Cancer Detection repository metadata says it was started 2026-05, last updated 2026-09.
My Skin Cancer Detection app uses Docker deployment metadata with app_port: 7860.
My project includes a chat-style AI interface.
My project includes live camera capture for skin images.
My project lets users reset the prediction or chat.
My project has a responsive user interface.
My project uses TensorFlow/Keras and Flask, with a JavaScript repository description.
The README describes an image-screening flow that preprocesses an uploaded or captured image, passes it to the ResNet50 model, and presents a Cancer / Non-Cancer result with confidence scores. It does not supply model performance metrics, dataset size, evaluation results, or details of confidence-score calibration, training, and validation.
My customer-churn-api repository is described as Python. Last updated 2026-09.
My AI_ChatBot repository is described as Flask AI chatbot using Gemini API with LangChain. HTML. Last updated 2026-06.
My movie-analysis-sql-project repository is described as SQL-based movie analysis with data cleaning, filtering, aggregation, and insights about ratings, genres, and top-performing movies. Last updated 2026-06.
My dipanshu-2106-Mobile-Sales-Analysis-Dashboard-PowerBI repository is described as Power BI mobile sales dashboard with interactive visualizations and KPI reports for sales trends, customer ratings, payment methods, and city-wise performance. Last updated 2026-06.
My power-bi-sales-dashboard repository is described as Power BI dashboard showing sales and profit insights. Last updated 2026-03.
My eye-blink-control repository is described as Control presentation slides using eye blinks with OpenCV and MediaPipe. Python. Last updated 2026-03.
I built DataPilot AI as an AI-powered data analysis platform on my GitHub. It is intended to take a dataset through cleaning, analysis, modeling, forecasting, and reporting, reducing repetitive analyst work. The project is also part of my project experience described on this page.
My DataPilot AI project is an AI-Powered Data Analysis, Machine Learning & Decision Support Platform.
The README says Data Analysts spend 70-80% of their time on repetitive tasks and says DataPilot AI automates these tasks in minutes.
The documented product modules are Data Quality, Data Cleaning, EDA, Dashboard, AI Insights, Ask Your Data, Root Cause, ML Lab, What-If, Recommendations, PDF Reports.
The listed forecasting methods are ARIMA, Exponential Smoothing, Moving Average.
My documented frontend stack is React 18, Vite, Tailwind CSS, Recharts, Lucide React, Framer Motion, Axios.
My documented backend and analysis stack is FastAPI, Pandas, NumPy, Scikit-learn, XGBoost, LightGBM, Statsmodels, FPDF2, Groq API.
The included sales demo dataset has 1,000 rows and 16 columns, with Sales, Revenue, Profit, and Customer ratings.
The README lists CSV, XLSX, and XLS support, a maximum file size of 100MB, and a minimum of 10 rows.
The repository is a Python project on @dipanshu-2106's GitHub, and its documented project structure includes frontend/, backend/, data/, and README.md.
The documented data API routes are POST /api/upload/file; GET /api/profile/{id}; GET /api/clean/preview/{id}; POST /api/clean/apply/{id}; GET /api/eda/{id}; GET /api/insights/{id}; POST /api/query/{id}.
The documented modeling and reporting API routes are GET /api/ml/columns/{id}; POST /api/ml/train/{id}; POST /api/forecast/{id}; POST /api/whatif/{id}; POST /api/rootcause/{id}; GET /api/recommendations/{id}; GET /api/report/{id}.
The documented development requirements and ports are Python 3.12+, Node.js 18+, backend port 8000, frontend URL http://localhost:5173.
```
GROQ_API_KEY=your_groq_api_key_here
LLM_MODEL=groq/compound
APP_NAME=DataPilot AI
```
The repository metadata says it was started 2026-09; last updated 2026-09.
The project is published under the MIT License — Free to use and modify.
The README presents the project's interview value as Python, machine learning, data analysis, GenAI, frontend, API design, software architecture, data engineering.
The README's setup sequence is to clone the repository, prepare the Python environment and backend configuration, start the FastAPI server, install frontend packages, and run the development server. The demo workflow proceeds from upload and data-quality review through cleaning, EDA, AI insights, natural-language queries, model training, forecasting, and PDF export. The supplied material documents the features and routes but does not report measured performance or model accuracy, and leaves implementation details such as cleaning rules, security controls, and request/response schemas unspecified.
I am based in Indore, Madhya Pradesh, India. My headline is “AI & Data Analytics | Python | SQL | Power BI | ML | DL | LLMs | AI Agents | Workflow Automation | n8n | LangChain | APIs & Integrations.” I turn data into insights and build practical AI/ML solutions using Python, SQL, Power BI, and automation tools like n8n. I am a B.Tech CSE student at AKS University and am looking for internship and entry-level roles in Data Analytics, AI/ML, and Automation.
I held the role Data Science Intern at SWYNEX Technologies (Sep 2026 to Sep 2026), Satna, Madhya Pradesh, India.
I held the role Data Analytics Intern at InAmigos Foundation (IAF) (May 2026 to May 2026), India.
I completed the Data Analytics Job Simulation at Deloitte (Apr 2026 to Apr 2026), India.
I studied at AKS University, Satna (M.P.): B.Tech (CSE) - AI&DS, Artificial Intelligence (Sep 2024 to Jun 2028).
I completed 12th at Naveen Saraswati Gyan Mandir H.S. school: 12th (Jul 2023 to Feb 2024).
I completed 10th at Naveen Saraswati Gyan Mandir H.S. School: 10th (Jul 2021 to Feb 2022).
I built UPI Fraud Detection: Anomaly detection system on 10,000 synthetic transactions (best F1 of 0.92 with Isolation Forest), served through a FastAPI service with a live Streamlit dashboard.
I built Forex Trading Signal AI Agent (ZeTheta Algorithms): Built automated buy/sell signal generation using Random Forest, LSTM, and Reinforcement Learning, with Power BI and R Shiny dashboards to track backtesting accuracy.
I completed in the Top 1%.
I ranked 2nd on the leaderboard during my InAmigos Foundation work.
My projects also include DataPilot AI, a platform that automates data profiling and ML model training with SHAP explainability, deployed with Docker on Hugging Face Spaces. In the SWYNEX internship, I worked on Titanic dataset preparation and cleaning, exploratory data analysis, baseline machine-learning models, and an end-to-end data science case study. At InAmigos Foundation, I researched NGOs across Education, Healthcare, Women Empowerment, Environment, and Animal Welfare, and created a structured database using Excel and Word with AI-assisted bonus analysis. The Deloitte simulation involved business-scenario analysis, forensic technology tasks, and data interpretation.
- My skills and tools include AI automation, Python, SQL, Power BI, Scikit-learn, TensorFlow, LangChain, n8n, FastAPI, Streamlit, Docker, Flask, reinforcement learning, LSTM, predictive modeling, Random Forest, business intelligence, data science, EDA, data cleaning, GitHub, Git, statistics, DSA, Microsoft Office tools, Oracle Database, PL/SQL, Microsoft SQL Server, deep learning, OOP, Matplotlib, and NumPy.
My GitHub is github.com/dipanshu-2106, and my email is 21dipanshugupta@gmail.com.
I built UPI Fraud Detection to explore catching suspicious UPI payments as they happen. The project compares anomaly-detection approaches on synthetic transactions and serves a model through an API and dashboard. Its data and performance are demonstrations, not evidence of real-world fraud-detection accuracy.
My UPI Fraud Detection repository is a Python project on GitHub; started 2026-10, last updated 2026-10.
My UPI Fraud Detection synthetic dataset contains 10,000 transactions, 500 users, about 2% fraud.
I engineered 7 engineered features per transaction, including amount z-score versus the user, hour, new device, new city, transactions in the last hour, and time since the last transaction.
In my project, is_fraud is used only for evaluation; no model is trained with it.
My reported results were IQR: flagged 610, precision 0.187, recall 0.57, F1 0.281, TP 114, FP 496, FN 86, TN 9304; Isolation Forest: flagged 200, precision 0.92, recall 0.92, F1 0.92, TP 184, FP 16, FN 16, TN 9784; Rolling Z-score: flagged 499, precision 0.18, recall 0.45, F1 0.258, TP 90, FP 409, FN 110, TN 9391; Ensemble: flagged 376, precision 0.316, recall 0.595, F1 0.413, TP 119, FP 257, FN 81, TN 9543.
My best reported metrics were precision 0.92, recall 0.92, F1 0.92 with Isolation Forest.
My evaluation limitation is that Isolation Forest is trained and evaluated on the same data; reported results are not a real-world estimate.
My detection methods are IQR per user, Isolation Forest, rolling z-score, and ensemble of at least 2 of 3 methods.
In this experiment, my ensemble was weaker than Isolation Forest because two of its three methods look only at the amount.
My API routes are POST /check, GET /health, GET /recent, GET /alerts.
My service provides LOW, MEDIUM, HIGH risk levels with plain-language reasons; every HIGH-risk transaction is logged, shown in dashboard Alerts tab, and shown as a popup.
My dashboard features are live feed, filters, trend charts, alerts tab, method comparison.
My project tech stack is Python, pandas, NumPy, scikit-learn, matplotlib, FastAPI, Uvicorn, Pydantic, Streamlit, joblib.
My repository structure includes api/, src/, app.py, simulate.py, run_all.py, docs/.
I can configure optional Telegram, email, and SMS configured with environment variables that must never be committed.
My cloud runtime is app.py, Python 3.12; free apps sleep after about 12 hours without traffic and first visit after a break can take a minute.
---
From Dipanshu Gupta's second brain at agentsocialx.com/21dipanshu