Dr. Anthony Brew

Researcher, engineer, and technology leader with >20 years of experience across academia, startups, scale-ups, and enterprises. Experience leading and coaching individual contributors, cross-functional teams & managers, enabling small expert units (<10) and multi-department structures (>50) to deliver state-of-the-art AI-enabled product suites while minimizing technical and scientific uncertainty in tandem with continuous customer impact.

Overview

Hard Skills: Generative AI, Recommender Systems, Automated Content Moderation, Audience Targeting,Text and Image Processing, Classical Machine Learning & Deep Learning, Anomaly Detection/Statistical Process Control, Software Engineering & Distributed Systems,

Soft Skills: Research Management, Product Management, Cross-Functional Team Leadership, Systems Design, Strategic Planning, Program Management, Agile Methodologies

Recent Technologies: AWS:(Sagemaker, Lambdas, EventBridge, ECR/EC2, EMR), Machine Learning/Big Data ( PyTorch, Scikit-learn, Snowflake, Spark, Flink, Kafka, MLflow), Web-Stack (Play-Framework, Django, React, Ruby-on-Rails, Streamlit) Deployments: (Jenkins, GitHub Actions, SAM/Cloud-Formation, Terraform)

Experience

Director of Generative AI
Shutterstock - Dublin, Ireland
July 2023 - Present

Shutterstock is a two-sided marketplace that allows contributors to sell images to customers. Generative AI disrupts this business model.

I lead the technology teams who deliver safe, indemnified Generative AI content at Shutterstock via API and through user interfaces. We are building a profitable credits-based product by integrating multiple models from external vendors (e.g. OpenAI, Google, Amazon) and working with strategic partnerships (Databricks & Nvidia) to build foundational models for different modalities. Our Image generator achieves a 70%+ monthly retention rate and are rapidly gaining stock-photography market share. Our growth surpasses other significant media types on Shutterstock, this is supported by a cohesive UX integration into Shutterstock.com driven by the following machine learning strategies:

Generative Model Optimization: We developed automated prompt refinement and parameter selection techniques, improving conversion rates by 30% on vendor-supplied models.
Generative Model Recomendations: We implemented an automated text-to-image recommendation system that leverages user prompts and page context to select the best generators. This increased conversion rates by 20%+. We enabled this by doubling the speed of validation using a purpose built traffic routing system which enables scientists to deploy independently.
Safety and Indemnity: We launched automated moderation mechanisms that achieved ~100% recall in key brand safety areas (e.g., child safety) while maintaining high precision to avoid penalizing non-violating users.

VP of AI & Engineering
Streetbees :: London & Lisbon (Remote)
Oct 2021 - July 2023

Streetbees is a Series-B funded AI SaaS service that helps brands like PepsiCo and Unilever understand market trends by aggregating insights from over 3.5 million customers. This is done via an AI-powered mobile chatbot platform that drives open-form surveys, automates survey response annotation, and delivers machine learning-driven market insights.

I led the technology organization (>50 members), divided into three stream-aligned teams: acquisition, annotation, and reporting. Each team included engineering managers, software developers, data scientists, DevOps, QA, and data analysts. I was responsible for the technical and scientific strategy across the organization, ensuring each team could operate autonomously.

Notable achievements under my leadership:

Aquisition: Reduced human moderation time for fraudulent image submissions by 75%. Our algorithm automatically identified over 80% of near-duplicate image submissions and screen captures as part of a review recommendation engine.
Annotation: Cut manual coding costs by over 60% by upgrading a legacy text-only classification system. Our cross-modality deep learning model increased automated code discovery by over 300%, removing the need for extensive manual annotation in primary survey domains (such as beauty).
Reporting: Fully automated the survey curation, delivery, and research report creation using GPT-3. This innovation cut the end-to-end client delivery time from over two weeks to days. This has ultimately led to a company pivot and the launch of a new product, Streetbees-X.

Head of NLP Centre of Excellence
Zalando :: Berlin (Remote)
Jul 2020 - Oct 2021

Zalando established four Centres of AI Excellence to drive innovation across areas such as NLP, Image Processing, Econometrics, and Forecasting. These centres enabled and seeded teams across Zalando with cutting-edge technologies.

I set up and led the NLP Centre of Excellence. I was responsible for the technical, product, and scientific direction of the centre. The team, consisting of 10 senior researchers, product managers, data engineers, and a data analyst, incubated technologies and supported the formation of new product-focused research teams across Zalando.

Key accomplishments include:

Platform Capabilities: Developed a GDPR-compliant, multilingual customer feedback search and message routing tool. This platform ingested data from 18 languages across 40 customer touchpoints, consolidating all customer feedback into a single platform.
Content Moderation: Incubated an automated moderation team for product safety issues flagged by the customer returns process. The tool, based on our state-of-the-art research, achieved over 95% recall in flagging unsafe products, expediting their removal with human-in-the-loop escalation to the manual inspection team.
Product-Led Scientific Advances: Developed novel text clustering techniques for identifying new categories of customer feedback and product risks. We also pioneered a zero-shot classification technique using transformer architecture and contributed to the Flair open-source NLP project, originally developed by Zalando Research.

Supervisory Board Member
Zalando SE
May 2018 - Jun 2020

I served as a member of Zalando’s 8-seat supervisory board, working alongside key shareholder representatives including Cristina Stenbeck (Kinnevik), Anders Holch Povlsen (CEO of Bestseller Group), Kelly Bennett (CMO of Netflix), Mariella Röhm-Kottmann (KPMG), Alexander Samwer (Rocket Internet), and fellow Zalando employees Konrad Schäfers and Beate Siert.

The board was responsible for fiduciary oversight of Zalando. This included making decisions on significant policy changes, investments, divestitures, executive promotions, remuneration, and managing relationships with strategic investors and key brand partners.

Head of Customer Data Platform
Zalando
Jan 2016 - Jul 2020

Zalando, Europe’s largest fashion e-commerce platform, expanded its technical workforce in 2015 by opening new technology hubs in Dublin and Helsinki.

I was a founding member of the Customer Data Platform and progressed from Senior Scientist to Head of Applied Science. I led a team of 25, including managers, applied scientists, engineers, and product managers, organized into three teams: 1) Customer Preference Recommendations, 2) Customer Price Predictions, and 3) Cross Device Graph. We integrated services across Zalando’s main business units: Lounge, Markets, Zalon, ZMS, and Fashion Store.

Key accomplishments include:

Customer Preference Recommendations: Our fashion attribute recommendation services, including category, brand, and price, powered customer preference collection, recommendation carousels, personalized navigation, gift card recommendations, newsletters, and offsite targeted advertising. These integrations drove double-digit improvements in A/B tests over existing solutions.
Audience Targeting: Our predictive audience generation tools, used by ZMS, Zalon, and Markets, achieved up to 40x performance over legacy systems in early campaigns which externally drove client retention (ZMS) and internally enabled deeper market penetration (Zalon & Markets).
Cross-Device Graph: Our cross-device graph improved advertising attribution by replacing the marketing team’s user association model. This led to a 40% increase in attributed sales significantly enhancing Zalando’s marketing performance. We also integrated this system into Zalando’s A/B testing platform, reducing test noise by double digits for cookie-based tests.

Lecturer : Distributed Systems
Technical University of Dublin
2012 & 2013

Lecturer for final year software development course and masters transistion course focused on distributed systems. Course practical sessions facilitated in both Java and Python. Covering core modern software components such as No SQL (Cassandra), Map Reduce (Hadoop), distributed queuing (Kafka), network communication protocols (e.g. TCP and UDP).

Senior Applied Scientist
IBM
Apr 2012 - Jan 2016

I was the Lead Applied Scientist in IBM’s SmartCloud division, responsible for inventing, optimizing, and delivering machine learning software to enhance IT network monitoring. I was initially hired to drive machine learning-enabled metrics management and later helped form a new team focused on event analytics.

Key accomplishments include:

Event Analytics: As part of a two-person discovery team, I co-developed new event analytics capabilities that helped protect IBM from immediate competitive threats. With market traction, I expanded the team to 14 contributors across the USA, UK, and Ireland. We covered design, UX, engineering, and applied science. I co-invented 16 patents in event analytics, reducing client workloads by 30% and decreasing Mean Time to Repair (MTTR) by 10%.
Metrics Management: I led a major redesign of the metrics anomaly generation product, improving disk I/O by 1000x (from 3.5 GB to 3.5 MB per 5-minute interval) and processing speed by 40x (from 5 minutes to 7 seconds per interval). The redesign also cut memory use by 2.5x (from 70 GB to 20 GB) and enabled independently testable machine learning extensions. This doubled the system’s detection methods, added memory leak detectors, and cut false alerts by 50%.

Data Scientist / Software Engineer
Swrve
June 2011 - Mar 2012

Developed Swrve’s SaaS A/B testing engine and implemented tracking metrics based on tracking event streams. Implemented front-end UX in Ruby & JQuery, backend in Java with a Redis, Cassandra, and MySQL storage layer, hosted on Amazon EC2.

Formal Education:

University College Dublin, Post Doc, Machine Learning :: Comercialization; 2010 - 2011
University College Dublin, Doctor of Philosophy, Machine Learning :: Speaker Verification 2006 - 2010
Trinity College Dublin, Master of Science :: High Performance Computing, 2002 - 2003
Trinity College Dublin, Bachelor of Arts :: Mathematics, 1998 - 2002

Projects & Courses:

The Automated Anglican: A automated digitization of my fathers history of sermons leveraging Google Tesseract for OCR, GPT for text correction & summarization and Dalle-3 for image creation served as a Jekyll website build and hosted by Github. Also made available as an OpenAI custom GPT for sermons-on-demand.
Stanford XCS224N - Natural Language Processing with Deep Learning: Professional qualification from Standford attained in order to bring my 2009 PhD knowledge up to the current state-of-the-art.