Hi! I'm Shreyan,
Projects
A showcase of my research, academic, and personal projects across various fields:
- Developed 3D Rubik's Cube visualizer using Quaternions.
- Implemented MCTS algorithm with a Deep RL network.
- Solved 71% of 9-scramble cubes, averaging less than 1 second per cube.
GPT Document Proofreader App
- Customizable prompts with English language variant support.
- Real-time async text processing with GPT-4o.
- Efficient context window management for handling large documents.
Volume Breakout Trading Algorithm
- Real-time volume breakout scanner with YFinance data integration.
- Interactive Plotly charts with volume analysis.
- Developed Streamlit dashboard with customizable strategy parameters.
MediLoRA: Mistral-7B fine-tuned for Medical Q&A with QLoRA
- Boosted PubMedQA and MedQA accuracy by 20%+.
- Matched SOTA on MMLU-Medical using minimal data.
- Demonstrated efficient fine-tuning with QLoRA technique.
Image Outpainting with Self-Attention GAN
- Implemented Self-Attention GAN from scratch.
- Reduced Reconstruction Loss by 3%.
- Improved Discriminator Loss by 2%.
Unfooling LIME and SHAP with superior sampling
- Counteracted adversarial attacks on explainability techniques.
- Superior sampling using Tree Ensembles and RBF Generators.
- Robust sampling using Variational Auto Encoders and CT-GANs.
Black–Scholes Option Pricing Using Machine Learning
- Outperformed the Black–Scholes Model using MLP and LSTM.
- Also implemented XGBoost and SVM for comparison.
- Published in the Proceedings of ICDSA 2022.
Causal Discovery in Lagged Time Series
- Evaluated PCMCI and FullCI algorithms for causal link detection.
- Demonstrated PCMCI's effectiveness in high-dimensional data.
- Showed PCMCI's robustness against extraneous variables.
Selective Lossy Image Compression for Autonomous Systems
- Self-devised algorithms for K-Means, Genetic Algorithm, and DCT.
- Integrated SOTA CNN-based models - CenterNet and Faster R-CNN.
- Presented in STSIVA 2021.
Experience
From research labs to dynamic tech startups, I've transformed data into insights and ideas into solutions. Explore my professional journey and discover the experience I've gained at the intersection of technology and real-world challenges.
Business Intelligence Engineer
Amazon, Austin, Texas
June – Present
2025
- Built Redshift ETL pipelines processing terabytes of customer data daily.
- Developed QuickSight dashboards and SQL queries for leadership BI metrics.
Machine Learning Engineer
Prompt Inversion, Dover, Delaware
September – December
2024
- Developing LLM-powered API for intelligent task automation.
- Optimizing codebase for scalability and efficiency.
- Building a modern web platform using Next.js, React and TypeScript.
Analytical Scientist Intern
FICO, San Diego, California
June – December
2023
- Fraud Detection Neural Networks.
- Adaptive Time Series Algorithm.
- Real Time Monitoring System.
- Terabyte-scale Datasets.
- Data Simulations and Fraud Schemes.
Research Engineer
Collablens, Haryana, India
February – August
2022
- AI station for automating drop tests.
- Contract to deploy in 50 factories.
- Computer Vision System for defect detection.
- $200,000 in investments.
Machine Learning Intern
Hypertechpreneurs, Haryana, India
May – December
2021
- Vehicle Damage Detection Model.
- Integration with mobile app and website.
- OCR and Document Analysis.
- $50,000 in investments.
Lead Research Fellow
Delhi Technological University
September – May
2021
- Option pricing using LSTM.
- Outperformed the Black–Scholes Model.
- Published research in the Proceedings of ICDSA 2022.
Research Assistant
Delhi Technological University
February – August
2021
- Selective Lossy Image Compression.
- CNN-based Object Detection.
- Published research at STSIVA 2021.
Contact
For any inquiries, drop me a message using the form below, and I'll get back to you shortly!