Alec Stashevsky

About Me

Technical leader with a track record of shipping machine learning / AI solutions at scale. I love building things people use, and I’m passionate about interdisciplinary research and application. Currently, I am leading teams working at the crux of computer vision, natural language understanding, and graph machine learning.

When I am not thinking about graphs and transformers you can find me snapping shots on my Yashica T4 Super D, hiking, and making pottery.

Interests
  • Graph Machine Learning
  • Search and Recommendation
  • Inequality and Development
  • Occult Economies
  • Place-making, Migration, Mobility
  • Spatial Politics
  • Anthropology
  • Decarbonization
Education
  • B.A., Mathematics-Economics

    Reed College

Publications

Domestic Remittance in China: Rural—Urban Migration’s Trail of Inequality
Domestic Remittance in China: Rural—Urban Migration’s Trail of Inequality

This work examines the flows of rural-to-urban migration in China with a focus on the trail of remitted cash and its role in rural village income inequality. It is the first to decompose rural household income inequality by income factor component using all 5 waves of rural household data available from CHIP (and RUMiC) surveys. Repeated cross section data is used to examine trends of the rural income profile and income inequality. Rural household income inequality is decomposed by factor component through the Gini index, the constant of variation, and the half-squared coefficient of variation. This thesis focuses the lowest and most vulnerable strata of the rural household income distribution with attention to different income sources’ capacity to alleviate poverty, situated within the broader context of Chinese liberalization. Migrant worker remittance flows are framed as spatial links whose proliferation is a co-production of an increasingly liberalized and competitive setting both in rural villages and urban centers of China. Remittance share regressions, analogous to Engel Curve regressions, are run to examine the differential impact of remittances across the rural income gradient. The statistical dispersion of remitted income is used as a proxy to illuminate the links between migration and a shifting gradient of rural mobility. Remittance income is found to have significant mitigatory effects on rural income inequality. Households in the 10th-50th percentile of the income distribution are found to have a significant dependence on remittances.

Experience

 
 
 
 
 
Fetch
Tech Lead Manager, Core Machine Learning
Oct 2022 – Present Los Angeles
Tech Lead Manager for core Document AI. We build ML/AI to power the Fetch app. Our systems extract information from +10 million receipts in real-time every day, and process over $150 billion in gross merchandise volume annually, and equivalent to the third largest retailer in the United States.

We build and deploy solutions touching computer vision, natural language processing, graph machine learning, semantic search, and entity resolution.

Lead partnerships with open-source and academic communities including Stanford University, Hugging Face, PyTorch, PyTorch Geometric, AWS SageMaker, and Streamlit.
 
 
 
 
 
Fetch
Machine Learning Scientist
Oct 2021 – Oct 2022 Los Angeles
Document AI and Graph ML
 
 
 
 
 
NERA Economic Consulting
Financial Economist
Sep 2020 – Sep 2021 New York City
Look inside the books of some of the largest financial institutions in the world to estimate damages and predict the performance of complex financial instruments responsible for high-profile banking crises, securities fraud, and market-manipulation cases.

Build probabilistic financial models using Markov chain Monte Carlo methods, random matrix theory, and time-series forecasting.

Identify, explain, and value litigation involving mortgage-backed securities (RMBS), collateralized debt obligations (CDOs), swaps, and other derivatives underpinning trillions of dollars in assets.

Provide evidence and expert economic testimony for Fortune 500 companies, SEC, DOJ, and FINRA.
 
 
 
 
 
The Cadmus Group
Economist
Nov 2019 – Sep 2020 Portland, OR
Design and evaluation of demand-side management programs, including a $600k+ randomized control trial on smart thermostat direct load-control.

Forecast electric vehicle diffusion, demand elasticity, and electrification for budgeting hundreds of millions of dollars under diverse energy industry clients’ management.

Design difference-in-difference models, demand-elasticity programs, and causal inference mechanisms to provide gold-standard reporting to regulators and operators responsible for most of the United States energy supply.
 
 
 
 
 
Kyrgies
Data Scientist
Aug 2017 – Oct 2019
Designed marketing mix models, audience creation, and targeted advertising solutions for a bootstrapped e-commerce startup.

Deployed recommendation systems and probabilistic engagement models (pCTR, pConversion) for targeted ads and notifications systems that led to 8% reduction in CAC.

Certifications

Coursera
Generative Adversarial Networks (GANs)
See certificate
Coursera
Machine Learning Engineering for Production (MLOps)
See certificate
Coursera
Machine Learning
See certificate
Coursera
Foundations for Big Data Analysis with SQL (with Honors)
See certificate
Coursera
Analyzing Big Data with SQL (with Honors)
See certificate
DataCamp
Introduction to Spark with sparklyr in R
See certificate
DataCamp
Support Vector Machines in R
See certificate
DataCamp
Introduction to Scala
See certificate

Tools I Use

Python
Python
Rlogo
R
snowflake
Snowflake (SQL)
pytorch
PyTorch
huggingface
Hugging Face
pyg
PyTorch Geometric
aws
Amazon Web Services
docker
Docker
git
Git
kafka
Apache Kafka
redis
Redis
airflow
Apache Airflow
mlflow
MLflow
sagemaker
AWS SageMaker
stata
Stata

Contact