Rishav Aryan

Rishav Aryan

AI / ML Researcher & Engineer

CV

Download CV (PDF)

Summary

Applied ML engineer with 1+ years of experience in predictive analytics, forecasting, and decision optimization. Skilled in Python, SQL, data visualization, and translating complex models into actionable business insights. Experienced in building end-to-end data pipelines, experimentation frameworks, and monitoring systems that support operational and strategic decision-making.

Education

George Mason University
Master’s, Data Analytics and Engineering
Aug 2023 - May 2025

Vellore Institute of Technology
Bachelor’s, Electronics and Communication Engineering
Jun 2019 - Jul 2023

Experience

George Mason University

Research Associate - Machine Learning & Reinforcement Learning
Jul 2025 - Present

  • Built a regime-aware reinforcement learning system for financial time-series under non-stationary market conditions using VAE + GMM latent state modeling and hierarchical PPO policies.
  • Improved portfolio robustness from a baseline Sharpe ratio of approximately 0.85-0.95 to 1.23, increasing alpha from +0.05 to +0.23 during regime shifts.
  • Developed risk-aware training and evaluation pipelines to support stable policy learning for deployment-oriented decision systems.

Accure Inc.

AI Engineering Apprentice
Jan 2025 - May 2025

  • Contributed to a six-month industry-sponsored capstone focused on AI-driven cybersecurity threat detection and automated incident response.
  • Built and tested LLM-powered threat analysis workflows using SecureGPT, AWS Lambda, and S3 for network log classification and mitigation recommendation.
  • Applied NIST SP 800-53 grounded prompting to improve threat explanation quality and support security compliance mapping.
  • Collaborated on evaluation across simulated traffic datasets, supporting high-accuracy threat classification.

Wall Street Quants

Quantitative Researcher
Jul 2024 - Oct 2024

  • Developed and backtested adaptive momentum and reversal strategies in Python across top crypto assets, improving annualized returns from 18% to 25%.
  • Calibrated RSI and moving-average-based trading logic across thousands of simulated scenarios, improving Sharpe ratio by 15%.
  • Built data-driven evaluation workflows for model selection, risk filtering, and performance monitoring in highly volatile environments.

Foxmula

Machine Learning Intern
May 2022 - Jul 2022

  • Built and deployed an end-to-end ML pipeline for employee dissatisfaction and promotion prediction using XGBoost, stacking models, TensorFlow, and AWS.
  • Automated analysis workflows, saving 10+ hours per week and enabling proactive HR decision-making through interpretable predictive insights.
  • Engineered tenure, performance, and skill-gap features, improving model precision to 90% and increasing accuracy by 15%.

Publications

  • Multimodal Event Detection: Current Approaches and Defining the New Playground through LLMs and VLMs
    Co-Author, NLDB 2025
  • Left Atrium Segmentation Using Deep Learning Model
    First Author, IEEE INDICON 2022

Independent Research Projects

NetGuard

Agentic AI automating threat detection

  • Built an agentic AI system for real-time network traffic monitoring that correlates traffic data with anomaly logs to detect and classify cybersecurity threats.
  • Designed a serverless AWS pipeline with Lambda, S3, and concurrent chunk-based processing to scale ingestion, analysis, and aggregation of network security data.
  • Integrated SecureGPT with NIST SP 800-53 Rev. 5 grounded prompting to generate threat severity labels, mitigation actions, and explainable security recommendations.
  • Automated incident response through Jira ticket creation and validated model performance with up to 94.5% accuracy and 0.93 ROC AUC.

ConfluBot

AI assistant for Confluence documentation

  • Built a RAG-based AI assistant for enterprise documentation using semantic retrieval and OpenAI-powered generation over structured Confluence content.
  • Developed ingestion, chunking, indexing, and retrieval workflows to turn internal knowledge bases into grounded, searchable context.
  • Designed a modular architecture for future integrations across internal search, Slack, and multi-source enterprise knowledge workflows.

Enough Thinking

Improving LLM reasoning efficiency

  • Built a post-training framework for improving LLM reasoning efficiency using GRPO and SEAL-style approaches, reducing unnecessary inference-time compute while preserving answer quality.
  • Evaluated reasoning behavior across controlled tasks and optimization settings, focusing on model efficiency, response quality, and scalable experimentation workflows.
  • Developed modular training and evaluation pipelines for modern GenAI systems, demonstrating hands-on experience beyond API-level application development.

Technologies

  • Programming & Software Engineering: Python, SQL, R, Git, GitHub, GitLab
  • ML/AI: Scikit-learn, XGBoost, Random Forest, PyTorch, TensorFlow, predictive modeling, model evaluation, feature engineering, GenAI, RAG
  • Data Engineering: ETL/ELT, data pipelines, data modeling, batch processing, SQL analytics, data warehousing
  • Deployment & Infrastructure: FastAPI, Docker, AWS (EC2, S3, Lambda, SageMaker)
  • Analytics & Visualization: Plotly, Matplotlib, Seaborn, A/B testing, experimentation, dashboarding