Transforming Ideas Into Reality

0%

Abhi Bhardwaj

Engineering production-grade AI/ML solutionsโ€”from deep learning models and RAG pipelines to scalable full-stack systems.

Scroll to explore

About Me

I'm a Software Engineer and Machine Learning enthusiast pursuing my Master's in Computer Science at George Washington University. With a strong foundation in Algorithms, System Design, and AI/ML, I'm passionate about building intelligent software solutions that solve real-world problems.

My technical journey spans full-stack development, machine learning systems, and MLOps. I've built production-grade fraud detection models with 96%+ precision, developed scalable e-commerce platforms handling 1000+ concurrent users, and created AI-powered financial analytics tools. From training neural networks to architecting cloud-based microservices, I focus on writing clean, efficient code that scales and delivers measurable impact.

Featured Projects

๐Ÿ”ฎ Flagship Project

Lumina AI

Visual Commerce Engine

Solving the $500B visual search problem in e-commerce. AI-powered engine that understands fashion through computer vision โ€” zero-shot object detection, style & vibe analysis, and semantic search with two-stage retrieval and cross-encoder reranking. Production-grade with CI/CD, Redis caching, and hybrid search filters.

OWLv2 CLIP Qdrant FastAPI Next.js 15 Redis Docker
Lumina AI Screenshot
7+ AI Models
Live HF Spaces
2-Stage Retrieval
Full Docker Stack
โœฆ Zero-shot fashion detection (OWLv2)
โœฆ Style & vibe analysis (CLIP)
โœฆ Cross-encoder reranking pipeline
โœฆ Hybrid search: vector + structured filters
โœฆ Live on Hugging Face Spaces
Featured
FraudGuard Screenshot

FraudGuard - MLOps Fraud Detection

Python XGBoost MLOps MLflow FastAPI

Enterprise-grade fraud detection system achieving 96.2% precision using XGBoost and LightGBM. Complete MLOps pipeline with MLflow tracking, CI/CD deployment, real-time drift detection, and FastAPI microservice with <50ms latency. Delivers $2.4M annual savings through 40% reduction in false positives.

Featured
TicketBlitz Screenshot

TicketBlitz - High-Concurrency Booking System

Node.js Kafka Redis WebSockets React

Live Production System deployed on Vercel & Render. Features a Real-Time Engineering Visualizer exposing Redis distributed locks, atomic PostgreSQL transactions, and Kafka event streaming. Engineered to handle 10k+ concurrent requests with zero race conditions.

Featured
QuantTradingBot Screenshot

QuantTradingBot - ML Backtesting Engine

Python XGBoost PyTorch Streamlit

Professional quantitative trading research platform and backtesting engine. Implements XGBoost and PyTorch LSTM models to predict market direction using 12+ technical features. Mitigates overfitting using anchored walk-forward cross-validation. Features Markowitz portfolio optimization, Monte Carlo risk simulation, and an interactive Streamlit dashboard.

Featured
AgentForge Screenshot

AgentForge - Deep Reinforcement Learning

Python PyTorch DQN Gymnasium Streamlit

Deep Reinforcement Learning agent that teaches itself to balance a pole through pure trial and error using Deep Q-Learning (DQN). Features Double DQN, experience replay, ablation studies, and an interactive Streamlit dashboard with live training, convergence charts, and gameplay videos. 30/30 tests passing.

Featured
SentiStock Analytics Screenshot

SentiStock Analytics

Python NLP Streamlit TextBlob

AI-powered stock analysis platform integrating real-time market data with NLP sentiment analysis. Processes 100+ financial news articles per minute, achieving 75-80% correlation to price movements. Interactive Streamlit dashboard with <2s data fetch times across 10,000+ stock tickers.

Featured
UrbanKart Screenshot

UrbanKart - E-Commerce Platform

React TypeScript FastAPI MySQL Vite

Scalable full-stack e-commerce platform supporting 1000+ concurrent users with <100ms API response time. Features JWT authentication, role-based access control, MySQL triggers for inventory automation, and type-safe React frontend with Vite optimization achieving <2s load time.

MoodTunes Screenshot

MoodTunes

Python AI/ML Recommendation System

Emotion-aware music recommendation system that analyzes user emotions and preferences to curate personalized playlists. Uses machine learning to understand mood patterns and suggest songs that match your vibe.

Featured
PadhAI Dost Screenshot

PadhAI Dost v2.0

Next.js 14 RAG Pipeline GenAI EdTech

AI-Powered Study Support System. Transforms static PDFs into interactive conversations and auto-generated flashcards using retrieval-augmented generation. Features intelligent context-aware chat, hybrid search, and real-time learning metrics.

Featured
WeatherNow Screenshot

WeatherNow - AI Weather Intelligence

Python PyTorch FastAPI Streamlit LSTM

AI-Powered Weather Platform combining real-time data with LSTM neural networks for next-day temperature prediction. Features GPS geolocation, 30-day historical analytics with Plotly visualizations, and interactive Folium maps. Glassmorphism-themed Streamlit dashboard with Docker containerization and 100+ city support across India/US/EU.

Work Experience

AI Engineer

JPMorgan Chase & Co.

Mar 2026 - Present

  • Designed and deployed Retrieval-Augmented Generation (RAG) applications using Python, LangChain, OpenAI GPT, and Pinecone, improving response relevance and reducing analyst research time by 11.4 hours per week
  • Developed and optimized machine learning models using PyTorch, Hugging Face Transformers, and MLflow, increasing document classification accuracy from 88.1% to 94.6% for financial compliance workflows
  • Built scalable REST APIs and AI microservices using FastAPI, Docker, Kubernetes, and AWS to support enterprise AI applications with secure and reliable integrations
  • Improved LLM performance through prompt engineering, semantic search, embedding models, and response optimization, reducing average API latency by 1.8 seconds across 1.6 million monthly requests
  • Collaborated with cross-functional teams to deploy production-ready AI solutions, automate business workflows, and ensure high availability, scalability, and security across enterprise platforms

Software Engineer

Optima Financial Services

Aug 2023 - Jul 2024

  • Designed and developed RESTful APIs and microservices using Java and Spring Boot, reducing average response time by 640 milliseconds across customer-facing banking applications
  • Built and optimized ETL workflows using Apache Kafka, PySpark, and SQL to automate data processing, reducing overnight batch execution by 2.9 hours
  • Developed data preprocessing and feature engineering pipelines using Pandas, NumPy, and PySpark, decreasing feature generation time by 31 minutes for recurring machine learning workflows
  • Automated application deployment using Docker, Git, and CI/CD pipelines, increasing release frequency from two deployments per month to seven without critical production incidents
  • Performed unit testing, integration testing, and code reviews, identifying and resolving 61 production-impacting defects to improve application reliability and software quality

BI Big Data & Analytics Engineer

Vodafone Intelligent Solutions

Aug 2022 - Jul 2023

  • Developed and maintained scalable ETL pipelines using AB Initio, BigQuery, SQL, and GCP to process over 500K daily customer transactions for analytics and reporting
  • Engineered 50+ behavioral, temporal, and aggregated features from large-scale datasets using Python, SQL, and BigQuery to support analytics and machine learning initiatives
  • Automated data quality validation and pipeline monitoring using Python and Unix shell scripting, reducing manual QA effort by 80% and improving data reliability
  • Optimized big data processing and query performance on GCP, improving data availability and reducing processing time for business-critical reporting
  • Streamlined software deployment by implementing CI/CD pipelines with automated testing and staged rollouts, reducing deployment time from approximately 4 hours to 30 minutes

Technical Skills

Languages

Python C++ TypeScript SQL R Shell Scripting

Machine Learning & AI

TensorFlow PyTorch Scikit-Learn NLP RAG LLMs Computer Vision Deep Learning OpenCV

Web Development

React.js Node.js FastAPI Streamlit RESTful APIs

Cloud & Big Data

Google Cloud Platform BigQuery Cloud Storage AB Initio Teradata

DevOps & MLOps

Docker Kubernetes CI/CD MLflow Git

Databases & Tools

MySQL MongoDB ChromaDB Vector Databases Pandas NumPy LangChain

Education

๐ŸŽ“

Master's in Computer Science

George Washington University

Aug 2024 - May 2026

๐Ÿ† SEAS Merit Award Recipient

Coursework: Design & Analysis of Algorithms, Machine Learning, Cloud Computing, Computer System Architecture

๐ŸŽ“

B.Tech. in Computer Science Engineering

SRM University

Jul 2018 - May 2022

Coursework: Data Structures & Algorithms, Probability & Queueing Theory, Data Mining & Analytics, Artificial Intelligence

Articles

Deep dives into neural networks, distributed systems, RAG pipelines, and production ML โ€” straight from the trenches.

โœจ Latest

๐Ÿง  The Math Behind the Magic: My Neural Networks & Deep Learning Journey at GWU

A deep dive into neural networks, backpropagation, gradient descent, and my academic journey exploring the mathematics that power modern AI at George Washington University.

Read Article โ†’
๐Ÿง 
13min read
DeepLearning
01

๐ŸŒค๏ธ Building WeatherNow: Can AI Predict Tomorrow's Weather?

How I built an AI weather intelligence platform using LSTM neural networks for time-series prediction with 92% accuracy.

Read Article โ†’
02

๐Ÿ“š RAG-Powered Learning Assistant: Turning PDFs Into AI Tutors

Building a RAG-powered study assistant that transforms static PDFs into interactive learning experiences with contextual chat.

Read Article โ†’
03

โšก Zero to One: Handling 10,000 Concurrent Users with Distributed Systems

Building scalable distributed systems with Kafka, Redis, and WebSockets for high-concurrency ticket booking.

Read Article โ†’
04

๐Ÿ” Zero to One: Building an Enterprise-Grade Fraud Detection System

How I architected a production-ready MLOps pipeline with 96% precision using XGBoost, MLflow, and FastAPI.

Read Article โ†’

Get In Touch

I'm always interested in hearing about new projects and opportunities. Feel free to reach out if you'd like to connect!