MBA candidate at UNC Kenan-Flagler, venture investor at Excelerate Health Ventures, and builder of AI-powered product tools.

I work on projects at the intersection of: • AI systems • product strategy • decision frameworks

• AI product experimentation and LLM applications
• Build-vs-buy decision frameworks for AI systems
• Automation tools that simplify knowledge management
• AI-driven discovery tools for product opportunities

A framework for evaluating whether product teams should rely on GenAI APIs or invest in building custom fine-tuned models.

As organizations adopt AI, one of the most common product decisions is whether to integrate existing LLM APIs or develop internal models tailored to specific use cases. This project explores that decision through a structured comparison of cost, latency, and deployment complexity across different approaches.

The framework analyzes trade-offs between API-based models and fine-tuned systems to help product teams make more informed build-vs-buy decisions when designing AI features.

This work is also being developed further with faculty at UNC Kenan-Flagler as part of a potential academic research paper exploring practical evaluation frameworks for AI product development.

Tech: Python, HuggingFace Transformers, PyTorch, RoBERTa-base (fine-tuned), OpenAI GPT-4o, Claude 3.5 Sonnet, Gemini 2.5 Flash, scikit-learn, pandas, NumPy, Google Colab

A tool that automates the discovery and prioritization of relevant product management roles.

The idea came from the challenge of tracking new job postings while balancing coursework, internships, and other responsibilities. New roles appear across multiple platforms every day, and manually checking them repeatedly is time-consuming.

This project experiments with using AI to analyze job descriptions, compare them with a candidate’s background and career goals, and surface the most relevant opportunities. The system runs multiple times per day to capture newly posted roles, ranks them based on fit, and organizes them for review.

The goal is to reduce the manual effort involved in job discovery while ensuring that high-quality opportunities are surfaced quickly.

Tech: React, Node.js, OpenAI GPT-4 API, Google Places API, Vector Embeddings, PostgreSQL, Vercel

An automation tool that exports course materials from a learning management system and organizes them into structured Google Drive folders.

The goal was to eliminate the manual process of downloading and organizing files after completing courses. For example, across ~37 courses this would normally require several hours of manually saving PDFs, slides, spreadsheets, assignments, and feedback documents.

This tool automates that workflow and creates a structured Google Drive archive of course materials. Keeping everything within the Google ecosystem also makes it easy to explore the content later using tools like NotebookLM to surface insights and revisit past coursework.

Tech: Python, Canvas LMS API, Beautiful Soup (web scraping), Selenium, pandas, automated data pipeline

Python, HuggingFace Transformers, PyTorch, Model Fine-tuning (RoBERTa, BERT), LLM APIs (GPT-4o, Claude, Gemini), Prompt Engineering, Vector Embeddings

scikit-learn, pandas, NumPy, ML Evaluation (F1, MCC, ROC-AUC), Statistical Analysis, A/B Testing, Data Visualization (Plotly, matplotlib)

Build-vs-Buy Frameworks, Cost-Benefit Analysis, ROI Modeling, Product Experimentation, Strategic Tradeoffs, User Research

Git/GitHub, RESTful APIs, Cloud Platforms (AWS, GCP), Database Design (PostgreSQL, SQL), Agile/Scrum

AI product management
Product strategy
AI systems evaluation
Automation and developer productivity

LinkedIn: https://www.linkedin.com/in/teja-padala
Email: tejaswar.padala@gmail.com