I like figuring out why people do what they do, then building something about it.

I'm Vasavi, a founder and product manager who tends to move between user conversations, messy ideas, product strategy and actually making things.

This is where I document what I've built, what I've learned, and the questions I'm still trying to answer.

Explore My Work

This is how I think, build, and learn.

StudioAI

An AI product exploring how marketing teams can create and manage multi-channel content from a single source of truth.

RoleFounder & Product Manager
Timeline2025 – 2026
Customer Discovery30+
MVP Beta Users10+
Customer Discovery MVP AI Research Product Strategy

DotRyder

Product strategy work for an emerging mobility platform, spanning customer journeys, competitive research, pricing, and go-to-market.

RoleProduct Strategy Consultant
Timeline2024 – 2025
DeliverablesJourney Maps, Research, Strategy
FocusGTM Strategy
Journey Mapping Competitor Analysis GTM Strategy Pricing Feature Prioritization

Growth Alpha

Building and scaling a business from the ground up

RoleCo-Founder
Timeline2023 – 2026
Output20+ Engagements
FocusGrowth
Growth Strategy Analytics Brand Product Leadership

Exploring Deep Learning for Malware Classification

Deep Learning for Malware Classification
Completed

Explored how different deep learning architectures perform when classifying malware at scale. I benchmarked CNN, RNN, and other architectures across a 200GB dataset, achieving 95%+ accuracy and learning how representation and data quality affect model performance.

PythonTensorFlowCNNLSTMFNN
  • Feature engineering matters more than model complexity.
  • CNN outperformed baseline models on image-like representations.
  • Dataset quality directly impacted generalization.

Can Computer Vision Replace Manual Attendance?

Computer Vision Automation
Completed

Built a computer-vision attendance system to replace manual classroom attendance, using OpenCV and Django to detect and record students in real time.

PythonOpenCVDjangoSQL
  • The biggest challenge wasn't detection—it was making the system reliable across different lighting and positioning conditions.

Understanding LLM Reasoning Through Prompt Engineering

LLMs, Prompt Design, Reasoning
In Progress

An ongoing experiment into how prompt structure changes the quality, consistency, and failure modes of LLM outputs. I test different prompting strategies and evaluate where small changes in instructions meaningfully change outcomes.

PythonLLM APIsPrompt DesignEvaluation
  • Structure in a prompt often matters more than length.
  • Small wording changes can shift reasoning quality significantly.
  • Evaluation frameworks matter as much as the prompts themselves.