Exploring the intersection ofAI, Finance & Entrepreneurship
With a strong foundation in AI/ML research, software development, and fintech, I thrive on solving real-world challenges through technology and innovation.
My Background
I taught self-driving cars to see at 80% autonomy. Now I'm teaching algorithms to see money. NYU CS + Stern FinTech dual threat who went from scoring 98.1% accuracy grading 8,000 papers with BERT to building VaR dashboards that cut risk insights by 40%. My Monte Carlo simulations run 10,000+ scenarios before your morning coffee gets cold.
Plot twist: Ranked #1 out of 2,800 students in programming. Then Microsoft's Senior Director taught me to break GLIDE models at 96% precision. Now I break market inefficiencies with sub-2s latency trading systems.
Current Arsenal:
- →Event-driven trading strategies with CVaR and Sortino ratios
- →Real-time risk monitoring that makes compliance actually smile
- →pix2pix GANs hitting 90% accuracy (because why not?)
- →McKinsey Forward alum improving $10M+ SME forecasts by 12%
I don't do leetcode. I do production systems handling 10K+ concurrent users at 95% uptime. The kind where milliseconds cost millions and 'good enough' is career suicide.
Seeking: Teams where Python meets P&L, where algorithms have attitude, and where the only thing moving faster than our code is the money it manages. Fair warning: I've optimized everything from SD-WANs to cGANs. Your technical debt doesn't scare me, it excites me.
Technical Expertise
Professional Journey
McKinsey Forward Program
Apr 2025 - Aug 2025
McKinsey & Company United States (Remote)
- Collaborated with a global cohort to develop an AI-driven financial risk analysis tool using Python, AWS Lambda, and DynamoDB, improving forecast accuracy by 12% on $10M+ SME financial datasets
- Enhanced team efficiency by 20% by applying machine learning models, AWS cloud services, and Agile methodologies to optimize financial data workflows and collaborative resolution of technical challenges
Research Assistant
Apr 2025 - Present
NYU Stern School of Business New York, NY
- Engineered an iOS app powered by AI/ML APIs with a Flask, achieving 90% accuracy in sub-2-second meal analysis
- Architected a MongoDB-backed system with robust API integration, sustaining 95% uptime for 10K+ concurrent users
Teaching and Research Assistant
Dec 2021 - Jun 2024
SRM's Directorate of Learning and Development · Chennai, India
- Developed an AI grading tool for 8,000+ student submissions with BERT and BM25, achieving 98.1% accuracy and boosting efficiency
- Created a mathematical model for testing validity, implemented as a parallel entity to traditional evaluations
- Introduced as a supplementary evaluation tool to enhance accuracy and scalability for evaluating large volumes of papers
- Ranked 1st out of 2,800+ students in both C Programming and Object Oriented Programming
Web Designing and Software Development Intern
Apr 2023 - Aug 2023
Launchr Tech Delhi, India (Remote)
- Transformed and maintained client websites; guided SaaS tools for e-commerce with SEO and AI features
- Automated feature selection process for ML models, reducing model training time by 30% and increased traffic by 20%
Industrial Research Mentorship
Mar 2023 - May 2023
Microsoft Redmond, WA (Remote)
- Conducted Generative AI research on GLIDE model under Senior Director, achieving 96% precision in image generation
- Designed and optimized cGAN models, improving image diversity 30% through feedback and training over 80k steps
- Enhanced generative model reliability by 25%, ensuring ethical generative art with high-fidelity outputs
Software Development Intern
Nov 2022 - Apr 2023
VCOM Technologies PVT LTD Mumbai, India
- Developed AI-based SD-WAN optimizer using Python and Logistic Regression to improve routing and latency
- Built WAN analytics tools with Flask and Pandas, reducing transmission costs by 15%
- Integrated real-time anomaly detection via Decision Trees, deployed with Docker on AWS, cutting downtime 18%
Machine Learning Research Team Member
Jan 2022 - Mar 2022
Blackbox Singapore (Remote)
- Crafted Bayesian cross-validation models using Python, PyMC3, and NumPy to enhance predictions
- Optimized model selection with BIC and posterior checks, cutting interpretation errors by 15%
- Improved accuracy to 80.83% with 10-fold cross-validation using scikit-learn, ensuring consistency
Academic Background
Featured Work
Quantitative Risk Metrics Dashboard
- Minimised risk insights by 40% by deploying VaR, stress tests, and correlation analytics
- Built comprehensive dashboard using FastAPI, React, WebSockets, NumPy, MongoDB, and Matplotlib with simulated and real-time data
- Enabled real-time risk monitoring and analysis for portfolio management
Monte Carlo Portfolio Simulator
- Simulated 10,000+ return paths for portfolio forecasting using multivariate models
- Implemented using FastAPI, NumPy, SciPy, Polygon API, and Plotly
- Computed Sharpe ratio, VaR, volatility, and risk-return distributions for comprehensive portfolio analysis
Event-Driven Trading Strategy Simulator
- Backtested earnings and macro events with sub-2s latency
- Built using FastAPI, Redis, NumPy, Tiingo API, and WebSockets for real-time execution
- Calculated CVaR, Max Drawdown, Profit Factor, and Sortino ratios for detailed performance analysis
AI-based Automated Descriptive Answer Evaluation System
- Created scalable NLP autograder using BERT and BM25s, scored 8,000+ papers
- Submissions Will be coming from 80,000+ submissions with 98.1% accuracy
- Built with Python, MongoDB, PyMC3, and FastAPI for statistical model selection
Computer Vision and Perception for Self-Driving Cars
- Road segmentation, object detection, and tracking for autonomous navigation
- Improved visual processing and safety, achieving 80% system closeness to fully autonomous self-driving
- Implemented using Python, OpenCV, TensorFlow, and YOLOv8
pix2pix - Image-to-Image Translation
- Devised and trained a pix2pix conditional GAN, achieving over 90% accuracy on image-to-image translations
- Optimized training to 15 seconds per epoch on a V100 GPU after 200 epochs
- Built with PyTorch, TensorFlow, and CUDA for high-performance computing
Bayesian Model Cross Validation Machine-Learning
- Created a Gaussian Naive Bayes Classifier to predict income levels with Bayesian validation
- Achieved 0.8083 accuracy, ensuring strong predictive reliability
- Implemented using Python, scikit-learn, and PyMC3
Additional Information
Get In Touch
Interested in collaborating or have a project in mind? Feel free to reach out through any of the channels below or use the contact form.
© 2025 Utsav Doshi. All rights reserved.