About the role
This is not a passive role. You will work directly on the AI systems powering a product heading to market - one where the AI operates within the tight constraints of a mobile phone environment, dealing with limited screen context, predicting user inputs, and real-time performance demands.
Your work will influence real product decisions and ship to users. For the right candidate, this role can expand into broader ownership across product quality and system design
What You’ll Do
Design and improve prompts for classification and extraction tasks
Run structured evaluation cycles (accuracy, error analysis, iteration)
Build datasets for testing and validation
Analyze outputs from real usage (logs, SQLite data, structured results)
Write Python scripts for testing, scoring, and comparisons
Identify failure patterns and propose improvements
Validate system behavior in iOS simulation environments
Document findings in a clear and structured manner
What We’re Looking For
We care most about problem-solving ability and a bias toward action — credentials matter less than evidence you can think clearly and ship things.
Currently pursuing a Masters degree in CS, AI/ML, Data Science, or related field with a Bachelor’s degree in CS
Experience shipping at least one production-grade mobile app
Strong proficiency in React Native, Expo, and iOS ecosystem
Python proficiency for scripting and data validation
Experience working with APIs and local databases (SQLite or similar)
Strong problem-solving and debugging skills
Experience or formal training with LLMs and prompt engineering
Clear and proactive communication
Availability for full-time (or near full-time) internship
Strong oral and written communication skills
Ability to analyze results and iterate systematically
This Role May Not Be a Fit If
You have no experience with mobile app development
Your experience with AI tools is limited to basic usage
All of your experience is from coursework only
You prefer well-defined assignments with clear answers
You are not interested in analyzing and debugging system failures
Work Timing - 7:30pm IST to 4:30 am IST
Or may be flexible with a half day overlap with US teams - 3pm to 12am IST shift maybe
About our US client.
PiggyBank Ventures is an incubator building AI-first consumer applications. We are currently building Nestora - an AI-powered mobile app that helps people manage communication intelligently. This is applied AI at the edge: limited context, small models, real-time performance, and messy data. Accuracy and reliability matter.
Our team includes repeat founders and operators with experience building and scaling products, along with backgrounds from top-tier universities and technical programs.
Originally posted on Himalayas