Kevin J. Dolan

Work

• 

  Amazon Sales Estimation Model

  • Data Science
  • Data Engineering

  As part of Pacvue's data team, I implemented a novel Sales Estimation Model, crucial for Amazon sellers, significantly enhancing its accuracy and capabilities. Initially tasked with transitioning a proof-of-concept model into production, I identified and rectified flaws in data collection and model evaluation. Leveraging XGBoost, I developed an alternative model delivering results competitive with significantly more expensive enterprise providers. Moreover, I engineered a robust orchestration engine, facilitating billions of daily inferences for widespread use across Pacvue's mainline products, notably Helium 10's Market Tracker 360 benefiting tens of thousands of customers.

  • 
    Python
  • 
    XGBoost
  • 
    dbt
  • 
    dagster
• 

  Amazon Review Analyzer

  • Principal Engineer
  • Artificial Intelligence (AI)

  At Pacvue's AI Labs, I spearheaded research into cutting-edge AI technologies, particularly Generative AI, and I created a pioneering Aspect-Based Sentiment Analysis tool using Python and OpenAI's GPT- 4 LLM, capable of extracting structured data from millions of Amazon reviews without manual tagging. This tool, surpassing traditional NLP methods, was able to perform open-domain unconstrained aspect-based sentiment analysis, topic classification, as well as demographic data extraction. I productionized the product for a pending public release, and advised the frontend team on functional design.

  • 
    OpenAI
  • 
    Python
  • 
    Kubernetes
• 

  Snowflake Data Lake

  • Technical Director
  • Data Engineering

  I led the implementation of the Snowflake Data Lake at Helium 10, consolidating disparate application databases into a unified data lake for enhanced analytics. Using Python, I developed a YAML-based system to ingest data from multiple engines, ensuring freshness every 15 minutes via Snowflake and AWS S3. I also implemented Git-based change management and collaborated on DBT setup for efficient data modeling and transformation. This system processes over 10 TB of data daily.

  • 
    Snowflake
  • 
    dbt
• 

  OrderMetrics

  • CTO
  • Data Engineering
  • Financial Analytics

  I was personally responsible for the ground-up re-implementation of OrderMetrics, after it was acquired by Assembly. The core was a Python-based engine for collecting data from several heterogeneous ecommerce data sources capable of multiplexing over 10k/sec rate-limit-aware API requests. Additionally, I developed a library of over 100 globalized financial accounting metrics, and worked with a team of 6 frontend engineers to develop the application, supported by a GraphQL Backend in Node. OrderMetrics as a brand was shut down, but the technology was integrated into other Assembly products.

  • 
    Python
  • 
    Node.js
  • 
    GraphQL
  • 
    Snowflake
  • 
    React
• 

  DeepSleep Stories

  • Rapid Prototyping
  • Writing

  As a personal project, I developed an end-to-end proof-of-concept automated pipeline for the generation and narration of bedtime stories, designed to be distributed as a free nightly podcast, as well as a premium app. As a related exercise, I developed a collaborative AI writing tool and used it to write an 18,000 Novella in the Young Adult category. Samples available upon request.

  • 
    OpenAI
  • 
    Python
• 

  AddressReport

  • Chief Technology Officer
  • Greenfield Development
  • Data Engineering

  I built a web application (previously: Rentenna) to crawl rental listings in New York City, and display them with an associated “Rentenna Score” that assessed the listing for a variety of factors. We later renamed the application AddressReport and focused on building a full report of rich data collected from several private and public sources. This company was acquired by OnBoard Informatics.

  • 
    Python
  • 
    JavaScript
  • 
    Kubernetes
  • 
    amazon aws
• 

  Union Square Linens

  • CTO
  • Greenfield Development
  • Machine Vision

  After acquiring an online retailer of wedding linens, I developed an e-commerce backend and taxonomy for millions of product combinations. As part of this, I also developed a machine vision technology that allowed the automatic projection of fabric colors and textures onto in-context photographs of wedding linens to dramatically reduce the cost of producing high-quality photographs.

  • 
    Python
  • 
    JavaScript
  • 
    NumPy
  • 
    SciPy
• 

  FranchiseHelp

  • CTO
  • Frontend
  • Call-Center Operations

  I inherited a Wordpress website, and performed a full re-write to turn it into a successful modern web application. I then served as the sole engineer for the product, eventually hiring and managing two engineers. This included a web application frontend, a content-management system, and a lead-mangement system complete with call-center operations.

  • 
    Python
  • 
    JavaScript
  • 
    Twilio SendGrid