now
Shifu Health
I’m a co-founder. We are building an AI-assisted health service that works with certified human experts to make useful exercise and nutrition support more accessible.
projects
Things I’ve built and researched.
Products change, experiments end, and research becomes papers. This is a record of what I worked on and what each thing was trying to do.
products
The company began as dotflo and changed shape several times. The list runs from what I am building now back to the first version.
experiment
Home-service operations
We explored voice and workflow tools for the day-to-day operations of home-service businesses. This was one of several focused tests between Supamill and Shifu.
launched
Supamill: workflow automation
A conversational workflow builder. It connected apps, research, and browser actions, then built and ran multi-step automations from a plain-language request.
prototype
Supamill: desktop automation
An earlier attempt at a general computer assistant that could carry out repetitive tasks across desktop tools.
prototype
Voice AI for health benefits
A voice agent for health insurers and third-party administrators. We used the prototype to test where phone-based automation could fit inside benefits administration.
earlier version
Flo
A research-first AI sales tool for healthcare and health-tech teams. It found relevant contacts, managed outreach sequences, and researched messages at scale.
earlier version
Zama
dotflo’s research agent for prospect research and outreach, built to work inside the tools a business already used.
first MVP
dotflo
The initial company and product name. The first MVP was an AI-assisted analytics tool; the name stayed while the product moved through research and sales automation.
research
Georgia Tech
DARPA GARD
I worked with Polo Chau’s group on machine-learning robustness under DARPA’s Guaranteeing AI Robustness Against Deception program. The larger Georgia Tech effort was led with Intel.
Georgia Tech
Wellness Technology Lab
I worked as a backend developer on technology for health equity and accessibility while completing my master’s degree.
project
Sentiment Search
An interactive system for exploring how public sentiment around a topic changes across social and news sources over time.
papers
2022
Adversarial Pairwise Reverse Attention for Camera Performance Imbalance in Person Re-Identification: New Dataset and Metrics
New camera-level evaluation metrics, a 38-camera dataset, and an adversarial attention method for reducing camera imbalance in person re-identification systems.
Eugene P. W. Ang, Shan Lin, Rahul Ahuja, Nemath Ahmed, and Alex C. Kot. IEEE ICIP 2022.
2022
AppAxO: Designing Application-specific Approximate Operators for FPGA-based Embedded Systems
A method for designing approximate arithmetic operators around an application’s accuracy, performance, and hardware constraints.
Salim Ullah, Siva Satyendra Sahoo, Nemath Ahmed, Debabrata Chaudhury, and Akash Kumar. ACM TECS, volume 21, issue 3.
2022
Classification of Sleep-Wake State in Ballistocardiogram System Based on Deep Learning
A contactless sleep-wake classifier built from ballistocardiogram sensor data using a multi-headed deep neural network.
Nemath Ahmed, Srivyshnav KS, Kumar Chokalingam, Madhusmita Rawooth, Gulshan Kumar, Gaurav Parchani, and Vibhor Saran. IEEE EMBC 2022.