I'm an Applied AI Scientist with 5+ years of experience building and scaling solutions across Agentic AI, machine learning, and optimization. With a Master’s in Data Science from Indiana University Bloomington (GPA: 4.0), I’ve applied my skills to solve complex real-world problems at scale.
Currently at Apple, I architect multi-agent systems and tool-calling Data Ops assistants using LangGraph and Model Context Protocol (MCP) to automate complex scenario modeling. I also build production LLM-driven autograders and anomaly detection models leveraging Apple Foundation Models (AFM) to ensure strict training dataset integrity. Prior to this at Toyota, I led projects generating multi-million dollar value, including a $150M+ Accessory Recommendation System and an Annual Planning Optimizer built with Gurobi.
Throughout my career, I’ve shipped full-stack ML systems using AWS, Azure, Databricks, PySpark, and Airflow. Lately, my passion lies in building self-improving GenAI experiences and context-optimization loops. Right here on this site, you'll find “Ask About Prateesh”, a custom chatbot that answers questions about my background, my projects, and even my personal interests. It learns from feedback, stores unknown questions, and improves over time. Go ahead, test it out!
If you're hiring or collaborating on Agentic AI, ML, or optimization initiatives, feel free to reach out—the chatbot knows where to find me 😉
👉 Or better yet,
Download my resume here or
Email me
Built and deployed a Generative AI chatbot hosted on this website
to answer detailed questions about Prateesh’s background — a
RAG-based system tailored for recruiters and hiring managers. Used
OpenAI’s GPT‑3.5‑turbo with custom prompt engineering, LangChain
for orchestration, and FAISS for vector search over ada text
embeddings. Stores unanswered questions for continoues
improvement. Deployed serverlessly via Cloudflare Workers with
CI/CD, logging, and auto-monitoring of unseen queries.
To Provide Cognitive search capability to search against database
like FDA and EMA (European medical agency) via a natural language
question and return relevant results in order to help with
accelerating regulatory submissions for Eli Lilly. Performed
abstraction based Natural language generation methods like T5
Transformer, GPT-2 Algorithm and BART Transformer
PixelGram is a distributed photo-sharing app built with a
microservices architecture using REST APIs and RabbitMQ for
asynchronous communication. It supports features like image
uploads, user authentication, and activity feeds. The system is
containerized with Docker and deployed via Jenkins pipelines on an
OpenShift cluster for scalability and modular development.
In Banking and finance sector, the term Blockchain is frequently
heard. Each block contains a cryptographic hash of the previous
block, a timestamp, and transaction data. This project is an
effort to give bank the ability to access a single source of
information and also allows them to track all documentation and
validate ownership of assets digitally, as an unalterable ledger
in real time.
