Open to Internships & Projects

Hi, I'm Keshav Agrawal

Data Science & AI student who builds |

I'm an undergraduate at CHRIST University, Delhi NCR, and I like turning messy data into answers people can act on: pull it with SQL, wrangle it in Python, dig for the story, and put it on a chart that actually makes sense. If it involves data, a tricky question, and a stubborn bug — I'm in.

See My Work Say Hello
keshav_profile.py 
class Developer:
    def __init__(self):
        self.name = "Keshav Agrawal"
        self.education = "B.Sc. Data Science & AI"
        self.university = "CHRIST University"
        self.gpa = 3.4 / 4.0
        self.focus = ["Data Science", "Web Dev", "AI"]

    def get_status(self):
        return "Turning raw data into working products"

keshav = Developer()
print(keshav.get_status())
# Output: Turning raw data into working products
GPA 3.4 Academics
4 Shipped Projects
01 — Stack

The data analyst toolkit

From pulling data with SQL to wrangling it in Python and putting it on a chart — the languages, libraries, and tools I reach for to turn raw data into answers.

Programming Languages

  • PythonAdvanced
  • SQLAdvanced
  • JavaScriptProficient
  • PHPIntermediate
  • HTML / CSSAdvanced
  • RIntermediate

Web Development

  • HTML5 / CSS3Advanced
  • Vanilla JavaScriptProficient
  • PHP (Vanilla)Intermediate
  • FastAPIProficient
  • RESTful APIsAdvanced
  • Chart.jsProficient

Backend & Databases

  • MySQLProficient
  • FastAPI BackendProficient
  • Schema DesignIntermediate
  • User AuthenticationSecure
  • Session ManagementStateful

AI & Deep Learning

  • LSTM Time-Series ForecastingApplied
  • Computer Vision (CNN)Applied
  • LLM Integration (Ollama)Hands-On
  • Sentiment AnalysisApplied
  • Model ValidationWalk-Forward

Data Analysis & Visualization

  • PythonAdvanced
  • Pandas & NumPyAdvanced
  • Scikit-Learn & PytorchProficient
  • Matplotlib & SeabornProficient
  • ExcelAdvanced
  • SQL (Advanced)Querying
  • Tableau & Power BIProficient
  • Statistical AnalysisStrong

Tools & Methodologies

  • Git & GitHubAdvanced
  • VS Code / JupyterAdvanced
  • ClaudeDaily Driver
  • Postman API TestingProficient
  • OOP PrinciplesStrong
  • Prompt EngineeringAdvanced
02 — Work

Things I've built

Four projects — the data, the models, the backend, and the pixels on top. See GitHub to download the repo.

SmartSIP – AI-Driven Dynamic SIP Accelerator

An intelligent web application that optimizes Systematic Investment Plans (SIP) using deep learning price predictions and real-time news sentiment.

FastAPI LSTM Chart.js Llama 3.2 HTML/CSS/JS
1.40% LSTM MAPE
80.3% Regime Accuracy
0.25x-2x SIP Multiplier

Technical Implementation

  • Frontend UI: Complete animated, responsive client utilizing Chart.js for interactive live Nifty 50 price tracking charts, a custom animated SVG market health gauge, and a 30-day sentiment trendline visualization.
  • AI-Driven Backend: Fused real-time LSTM market projections and automated news sentiment parsing into 3 FastAPI endpoints to recalculate dynamic monthly multipliers.
  • LLM Integration: Connected Llama 3.2 locally via Ollama to generate free plain-English rationales explaining monthly investment adjustments.
  • Validation: Verified model performance across 10 years of Nifty 50 data utilizing 5-fold walk-forward validation.

Personal Budget Tracker – Full-Stack Web App

A dynamic, secure expense logging application designed with relational data storage, transactional insights, and modern DOM rendering.

PHP MySQL Vanilla JS HTML5/CSS3 Auth
100% Ajax Driven
Secure Sessions
Relational DB Schema

Technical Implementation

  • Frontend Engine: Implemented pure Vanilla JavaScript dynamic DOM updates, allowing users to filter transactions and add ledger logs seamlessly without browser refreshes.
  • Backend & Logic: Structured robust relational query logic and secure session tokens in vanilla PHP supporting complete transactional CRUD activities.
  • Database Schema: Crafted a normalized MySQL database structure linking `users`, `categories`, `transactions`, and `budgets` with strict referential integrity.
  • Smart Alerts: Built custom notification flags triggered via CSS transitions when a user's category-wise expenses exceed their set limit.

FaceMask Detection – DL Computer Vision

A high-performance, web-based computer vision tool enabling real-time face-mask usage detection and compliance tracking.

TensorFlow/Keras CNN OpenCV Vanilla JS WebRTC
Real-Time FPS Feed
CNN Architecture
Secure Edge Run

Technical Implementation

  • Browser Interface: Configured a responsive client-side layout enabling seamless system Webcam stream access and static image file parsing overlays.
  • Bounding Box Rendering: Programmed real-time Canvas rendering loops displaying multi-face detection borders, confidence rates, and status tags (Mask / No Mask).
  • Deep Learning Pipeline: Co-trained a high-accuracy Convolutional Neural Network (CNN) leveraging data augmentation techniques inside TensorFlow and Keras.
  • OpenCV Processing: Handled camera frames and image standardizations, packaging the application for safety compliance runs inside workspaces.

Credit Risk Modelling – Loan Default Prediction

A binary classification model to predict loan default risks, featuring intensive exploratory data analysis and model comparison.

Python Pandas/NumPy Scikit-Learn EDA Random Forest
Optimal AUC-ROC
4+ ML Models Tested
Actionable Risk Scoring

Technical Implementation

  • Pipeline Architecture: Engineered a robust Python classification pipeline parsing borrower parameters (debt ratios, income stability, history, balances).
  • Feature Engineering & EDA: Utilized Pandas and NumPy for extensive out-of-distribution tracking, missing data imputations, and outlier thresholding.
  • Model Comparison: Evaluated Logistic Regression, Decision Tree, and Random Forest classifier performances against key recall matrices and confusion grids.
  • Risk Indicators: Built a scoring utility mapping probabilistic model weights to actionable risk indicators for borrower solvency reports.
03 — Timeline

Education & credentials

Where I've studied and the certificates I've picked up along the way.

July 2024 – Present

Bachelor of Science (Honors)

CHRIST (Deemed to be University), Delhi NCR, India

Major: Data Science and Artificial Intelligence

A rigorous undergraduate curriculum focusing on quantitative mathematics, machine learning theory, programming methodologies, and data analysis pipelines.

3.4 / 4.0 Cumulative GPA
May 2027 Expected Graduation
Completed: Jun 2026

Databases and SQL for Data Science with Python

IBM (via Coursera)

Hands-on course covering relational database design and advanced SQL — joins, subqueries, aggregate functions, and views — then querying live databases directly from Python with SQLite, Db2, and pandas.

Verify credential ↗
Completed: Jun 2026

Google Data Analytics Professional Certificate

Google (via Coursera)

Nine-course professional program covering the full analytics workflow — preparing, processing, analyzing, and sharing data using spreadsheets, SQL, Tableau, and Python — finished with a hands-on capstone case study.

Verify credential ↗
Completed: Mar 2026

Financial Markets

Yale University (via Coursera)

Taken purely out of personal interest — an overview of the ideas, methods, and institutions behind risk management and finance, including behavioral finance, debt, equity, and the role markets play in society.

Verify credential ↗
Completed: Jan 2026

The Complete Full-Stack Web Development Bootcamp

Udemy (Instructor-Led Curriculum)

Comprehensive hands-on training spanning layout styling, relational databases, responsive clients, server-side design, session tokens, and complete full-stack project building.

04 — Off-Hours

Beyond the keyboard

Things I follow and read about when nothing is due.

Financial Markets

I enjoy following how exchanges, risk, and human behavior shape prices. Completed Yale University's Financial Markets course (Coursera, 2026) purely out of curiosity.

Motorsports

I follow racing across the board — F1, IMSA, WEC, MotoGP, WRC — and live for the endurance classics at Le Mans and the Nürburgring. Strategy, tyre calls, and lap-time data are half the fun.

Data Storytelling

Turning messy datasets into visuals people actually understand — dashboards, charts, and the occasional over-engineered graph nobody asked for.

05 — Playground

Poke around the terminal

Ask questions or inspect my portfolio directly via a custom CLI simulator.

guest@keshav-portfolio: ~
Interactive
Welcome to Keshav Agrawal's Terminal Interface v2.0.0.
Type help to view available interactive commands and start exploring.

guest@keshav-portfolio:~$
06 — Say Hi

Get in touch

Have an interesting project, internship opportunity, or just want to connect? Send me a message!

Location Noida / Sector 93A, India