Close

Arman Dave

Software Engineer

Download Resume

About Me

I am passionate about creating beautiful products that meet user needs in a simple yet powerful workflow. I graduated from MIT with a Bachelors in Computer Science and Math, and I am currently a Software Engineer at Block.

Experience

Block (Square)

Software Engineering Intern

I worked on the Automation Machine Learning Infrastructure team. I created a first version of an embedding model that could share merchant representations across various merchant-product recommendation models, thereby transferring knowledge across models and condensing model inputs. In the process, I worked with and refactored several different merchant-product recommendation model codebases.

Meta (Facebook)

Applied Research Science Intern

I worked on the Ads Auction & Delivery Team with a focus on ecommerce. For my main project, I built a rule-based model to create a user segment on Facebook Marketplace and then ran an experiment to validate the usefulness of the model on over 10 million users. As a side project, I conducted exploratory work to validate the need to mitigate drop-shipping in the Marketplace ecosystem and created a proof of concept to use text-based clustering in addtion to image clustering to identify drop-shippers.

Evavi

Software Engineering Intern

I worked on the Supply Chain Forecasting team. My project involved building an Airflow pipeline to run various Facebook Prophet forecasting algorithms and writing the outputs to a SQL database.

Arches

Quantitative Trading Intern

I worked on the Commodities Options Desk. My projects involved explorations into Convolutional Neural Network methods to analyze satellite imaging data and options strategies to profit on natural gas futures.

Education

MIT

Sept 2018 - Dec 2022

Bachelor of Science in Computer Science

Bachelor of Science in Mathematics

Carroll Senior High School

Sept 14 - May 2018

Projects

Abstract Frame Query Language

Video data grows exponentially every year, yet video analytics remains an open problem. In this paper, we propose a query language - Abstract Frame Query Language - that enables a user to query a video corpus in a SQL-like syntax. For the backend, our system operates on tuples and uses YOLOv5 for object detection - in addition to allowing users to plug in their own custom detectors. In our repo, we implement AFQLite, a CLI tool to quickly run analytics on video files.

View My White Paper
Check Out My Repo

Professor Marvin

Wake word spotting is a common problem, often used to wake home assistants such as Alexa or Siri with the phrase "Alexa" or "Hey Siri". We propose a wake word detection algorithm that runs on a ultra low memory and low power device (less than 1 MB). Our system, Professor Marvin, has been implemented by compressing a keyword spotting Convolutional Neural Network (CNN) system through the use of quantization and knowledge distillation to reduce memory footprint and FLOPs. The compressed model has been tested, reaching accuracy nearly as high as the baseline, and works on a Raspberry Pi architecture and can be deployed to edge devices.

View My White Paper
Check Out My Repo

Exhaustive Trust Algorithm (ETA): Mitigating the Flood & Loot Attack of the Lightning Network

Lightning Network is a Layer 2 solution on top of the Bitcoin blockchain that that aims to facilitate exponentially faster transactions without sacrificing the trustless security of the blockchain. A security flaw of the Lightning Network known as the Flood & Loot Attack was discovered in 2020. We designed the Exhaustive Trust Algorithm (ETA) and proved that an implementation of ETA would make Floot & Loot unprofitable.

View My White Paper

Weakly Supervised Instance Segmentation

We propose a weakly-supervised method using a limited set of annotated images from the Common Objects in Context (COCO) dataset to improve the performance of FreeSOLO, a state of the art self-supervised instance segmentation model.

View My White Paper

Tennis

2016 UIL 6A Tennis Boys' Singles Champion

UIL is the league in which all public high school athletes in Texas compete. I earned Southlake's first tennis state title in the newly created 6A division (for the largest public high schools in Texas). It was our first state title in nearly twenty years.

View Title

2016 Capsher Texas Grand Slam Champion

The Capsher Texas Grand Slam is the most prestigious USTA tennis tournament held in Texas. The tournament spans the course of an entire week. I won both singles and doubles in the B16 under division.

View Title

Skills

Get in Touch