Archer Williams | AI advisory services

Unlock the power of your data using AI and modern software practices

Archer Williams partners with mission-driven organizations to harness the transformative power of AI and emerging technologies. In our rapidly evolving digital landscape, many executives recognize the need to modernize their technology stack and leverage data assets effectively. However, navigating the complexities of AI adoption can be daunting. We help bridge this gap, by aligning strategy, talent, and processes to drive innovation and value creation through responsible AI implementation. We help organizations realize efficiency gains, unlock data-driven insights, and future-proof their operations by seamlessly integrating cutting-edge AI solutions tailored to their unique needs and goals.

Services

Archer Williams provides high-impact advisory and interim leadership services to help organizations accelerate innovation and growth

AI-generated aurora borealis over a mountain landscape

AI Advisory Services

AI readiness assessment & due dilligence
Organizational design
Data architecture review and design
Vendor analysis
Data and ML governance
Tech leadership and program management

Engineering Leadership

Talent assessment and cultivation
Hiring strategy and recruitment process
Career ladder development and salary benchmarking
Operations
Agile enablement
Mentorship and training program development

Product Leadership

Solution discovery
Persona development
User journey mapping
User centered design
Roadmapping
SDLC management & execution

Projects

Restoring Coral Reefs

A coral farm operator wanted to streamline operations by leveraging computer vision to rapidly measure coral growth and health across hundreds of thousands of microfragments. I led a project to evaluate different models, and develop pipelines to support training and deployment. The client was able to demonstrate advanced cultivation analytics at scale enabling them to win a key contract for a new farm.

Anonymizing Clinical Trial Data

A large pharmaceutical company needed to comply with regulations governing the anonymization of participants in a skincare clinical trial. I led a project to detect identifiable features (tattoos, piercings, facial features, etc.) to mask out the identifiable information to enable to trial to proceed.

Reading Ultrasounds to Measure Cardiovascular Health

A provider of ultrasound scanning services wished to remove the time and effort required to process an ultrasound study of the carotid artery. I led a project to harvest years of historical study data, anonymize it, and train two models:

A model to detect and characterize plaques in sequences of transverse views of the carotid artery
A model to isolate and measure CIMT from longitudinal views of the carotid artery

By automating these two manually intensive processes, our client was able to reduce the time to generate a study report from days to minutes.

Reading MRIs to Predict Surgical Complications

A spinout company from a prominent research institution had a proprietary regression model which could predict the probability of complications from spinal surgery based on the geometric orientation of spinal discs (and other features). However, the process of measuring was manual, and subject to high variability, leading to high error in the model's predictions. I led a project to standardize the geometric measurements by training a computer vision model to reliably and accurately segment discs in MRI scans and derive the geometry from the segments, resulting in much more consistent and accurate predictions.

Enabling Structured Analytics on Unstructured Data

A software product company serving HVAC maintenance technicians had accumulated millions of unstructured and semi-structured service records spanning many years. They wished to normalize this data and extract key information from each record, such as make and model of equipment being serviced, root cause of the malfunction, and solution to the problem in order to drive predictive analytics as a value add to their product. I led the project to use a large language model and multiple prompting techniques (few shot, chain of thought) to extract key structured elements from the unstructured text. We used Spark to parallelize the execution of the algorithm across millions of rows to convert one column of unstructured text into many columns of structured data suitable for predictive analytics.

Reasoning Over Large Unstructured Datasets Using Knowledge Graphs

A large cement repair contractor had accumulated a wealth of knowledge about assessing and repairing massive concrete structures, but the knowledge was hidden away in decades' worth of site assessments and statements of work. The client wished to extract meaningful relationships out of this unstructured library to train new employees, and inform decision making and standard procedures for future projects. I led a project (pre-chatGPT) to use traditional NLP approaches to map documents into a human-defined ontology. We then implemented retrieval augmented generation to enable querying of the graph and retrieval of relevant context in human readable form. This project also laid the groundwork for linking structured data (e.g. project performance metrics) into the knowledge graph to enable novel forms of analysis.

De-Duplication of Customer Records Over Large Datasets

A private equity backed firm needed to clean up millions of rows of customer data after multiple acquisitions and years of neglect. I led a project to analyze the dataset and discover the business rules for deduplication. We then developed high-scale data cleansing routines to evaluate the business rules efficiently across the entire dataset to merge duplicate records. With cleaned customer data, processes that formerly took multiple days were achieved in minutes.

A Standard Environment for Behavioral Analysis

A principal investigator at a prominent research institution won a grant to study the impact of different interventions on couples where one member is afflicted with dementia. The study involves recording and aggregating sensor data from hundreds of homes to measure daily engagement and how it changes over time in response to specific interventions. She was struggling to aggregate sensor data from the homes in a consolidated environment where analysis could be performed in a repeatable manner. I led the project to equip her team with a data architecture to support ingesting, cleaning and analyzing the study data which serves as a model for repeatable research throughout the institution.

Data Architecture to Support Business Intelligence

I've led multiple projects to ingest data from multiple critical business systems (CRMs, HRIS, Accounting, ERP, and custom product databases) into a centralized data lake to support unified business analytics. I start with a detailed data dictionary for each data source, highlighting the key business entities and their relationships to other systems and entities. Then I build pipelines (or use off the shelf software) to ingest raw data into a bronze layer as a staging area for subsequent refinement and joining pipelines. With all of the data in one system being refreshed predictably and reliably, off the shelf BI tools can serve the company and truly democratize access to the data needed to support rapid awareness and decision making.

Hiring and Team Building

I have built high-performing multi-disciplinary teams at companies large and small. I have built a from zero to thirty in six months, while building other teams gradually over multiple years. And of course, I have inherited teams of various sizes. I bring a proven playbook of competency and scenario-based interviewing which enables the hiring team to make evidence-based hiring decisions while creating a delightful experience for candidates.

Developer Productivity

Low developer productivity impedes product development and destroys developer morale. Since my first management job, I have sought to maximize developer productivity. I have had to shift engineering culture at multiple companies to embrace automated testing. I have leveraged tooling for standardizing developer workstations (first Maven, then Docker, now DevContainers and cloud development environments). At eBay, I designed and implemented automated build and test infrastructure for mobile apps at massive scale to offload work from developer workstations into a custom datacenter, while simultaneously working with IT to improve developer hardware and contain/minimize corporate bloatware to reduce local compile time.

Advances in generative AI promise to supercharge developer productivity. I've been keeping a close eye on the ecosystem of tools for rapid onboarding, reasoning over large code repositories, and other productivity enhancements aimed at the craft of software engineering.

Performance Management & Compensation

Building a team is one thing, but keeping a team energized and engaged is a different challenge. I focus on consistently communicating the team's mission and how it maps to the success of the company. To make this tangible, I favor the Objectives & Key Results approach (OKRs). I have instituted and managed OKRs at three different companies, while driving public data-driven dashboards to enable rapid feedback.

Another important means of keeping teams engaged is to ensure each member understands where they are on their path to career development. I have created technical career ladders at both large enterprises and small startups. I favor pay transparency, so employees understand how their performance maps to different salary bands. This also helps ensure equity by not favoring employees who are simply good negotiators.

About

I am the owner of Archer Williams, and principal consultant. My interdisciplinary education in biochemistry and cognitive science at the University of Virginia instilled in me a deep fascination with how science and technology can solve complex problems. I put these analytical skills to work first as a software engineer at Intel and Barclays, then managing a team of developers on a molecular modeling product at Fujitsu BioSciences.

I transitioned into engineering management at Jive Software, scaling their professional services team through the company's successful IPO. I then moved into a product director role, leveraging my technical and business acumen to solidify Jive's position as a social business software leader.

I brought this unique blend of technical and product expertise to eBay. There, I built high-performing, cross-functional teams that drove mobile app innovation and pioneered the use of machine learning for anomaly detection in mobile telemetry data.

After taking Andrew Ng's influential Machine Learning course, I was inspired to dive deeper into the field of applied AI. I joined Synaptiq, a boutique AI consulting firm. Once there, I rapidly grew the team over fourfold while delivering transformative computer vision, NLP, and predictive analytics solutions for clients in a lead consultant role.

Now, as an independent consultant, I'm laser-focused on harnessing the power of artificial intelligence to drive breakthrough innovation. Whether it's revolutionizing climate tech, transforming healthcare, or improving government services, I'm dedicated to tackling high-impact projects that create tangible value for organizations and society.

Contact

+1 (503) 869-3546

info@archerwilliams.com

AI-generated landscape with mountains and river

https://www.linkedin.com/in/erskine/

Your Partner in AI-Driven Business Transformation