Aditya Mehrotra

A platform for building and backtesting cross-asset trading strategies in plain English.

• •Building an LLM-native backtesting platform: describe a trading strategy in plain English, get a validated, executed backtest and PDF report back
• •Three-LLM pipeline (generation, deterministic validation, report commentary) over a phased backtest engine with 46 technical indicators
• •Fine-tuned Qwen2.5-Coder-7B with LoRA to match Claude Sonnet on strategy-config generation: 98.2% holdout accuracy on a 26K-example dataset

Deterministic risk guardrails for AI trading agents, shipped as a Python library, MCP server, and Claude skill.

• •Correlation-aware overexposure detection that flags when multiple positions are effectively the same bet
• •Trade scoring, position sizing, portfolio heat, and stress testing exposed as agent tools
• •Composes with broker MCPs like Robinhood as a read-only advisory layer between agent and execution

Paper trading platform with virtual portfolios, live quotes, and trade history.

• •Simulates real market investing with virtual currency, live quotes, and multiple portfolios per user
• •OAuth sign-in with per-user accounts, saved portfolios, and full trade history
• •Built with a React and TypeScript frontend and a Spring Boot backend, on AWS EC2 with a MongoDB cluster

A Wordle solver published on PyPI with five pluggable solving strategies and a full benchmark suite.

• •Benchmarked all five strategies over every valid Wordle word, with entropy winning at 3.47 average guesses and a 100% solve rate on the classic answer list
• •Shipped as an installable CLI with CI and published releases

Leaf species classifier from photos, with auth and a monitored production deployment.

• •Custom-trained TensorFlow model that classifies leaf species from a photo at 86% validation accuracy
• •Uploads stored in S3, with user accounts and classification history in MongoDB
• •React frontend and Flask API, deployed on AWS and monitored in production

A multiplayer Monopoly web game with full gameplay mechanics

• •Developed comprehensive multiplayer game logic with property management and trading systems
• •Built interactive React/TypeScript frontend with real-time game state updates
• •Deployed Django backend on AWS EC2 for scalable multiplayer game hosting