Allan Ilyasov

Allan Ilyasov

AI/ML Engineer & Full-Stack Developer

About Me

I'm a Graduate Student pursuing an MS in Data Science and BS in Computer Science at St. John's University, graduating in May 2026 and May 2025 respectively. Currently working as a Graduate Research Assistant, I specialize in building AI-powered applications and scalable cloud infrastructure.

My expertise spans full-stack development, machine learning, and AWS cloud services. I've led the development of Codify AI, an AI-powered programming tutor, and won multiple hackathons for innovative AI solutions.

When I'm not coding, I'm contributing to Uncle Edik's Pickles, a startup I helped grow from a home-based operation to a national brand.

Technical Skills

Languages

PythonTypeScriptJavaScriptJavaSQLRSwiftPHP

Frontend

ReactNext.jsTailwind CSSRadix UI

Backend

FlaskDjangoSpring BootNode.jsREST APIs

AI/ML

LangChainLangGraphTensorFlowPyTorchScikit-learnAWS Bedrock

Cloud & DevOps

AWSDockerDynamoDBElastic BeanstalkLambdaAPI Gateway

Data Science

PandasNumPyPlotlySparkNLTKSpacyBeautiful Soup

Databases

MySQLMongoDBPineconeFAISSVector Databases

Projects

Diabetes Risk Prediction

Tested whether lifestyle or demographic factors are greater predictors of diabetes using a dataset of 230K+ patient records. Employed advanced data wrangling (imputation, encoding, outlier detection) and addressed class imbalance with SMOTE. Developed ML algorithms (Random Forest, Logistic Regression, KNN) with 5-fold cross-validation to predict diabetes risk.

PythonScikit-learnSMOTE+4 more

Codify AI

AI-powered programming tutor leveraging AWS Bedrock (Claude 4.5 haiku), achieving 50% response time improvement through optimized prompt engineering, and LangChain-based agentic workflows.

ReactAWS BedrockFlask+3 more

Time Series Forecasting for Financial Markets

Built ARIMA, LSTM, and GRU models for stock prediction (AAPL, NVDA, LYFT), achieving 64-82% RMSE reduction. Implemented grid search across 36 ARIMA configurations and built 2-layer LSTM/GRU architectures with dropout regularization. Conducted stationarity analysis using ADF tests and automated data extraction for 1,458+ trading days via yfinance API.

PythonTensorFlowStatsmodels+4 more