How do you get started building an AI team? Hire a unicorn that understands your business problem, translates it into ‘right’ AI building blocks, and can run implementations and operational deployments. Sounds easy! However, such unicorn sightings are extremely rare. Even if you find a unicorn, you probably can’t afford it.

Based on my experience leading Data + AI products and platforms over the past 20 years, a more effective strategy is to cumulatively support his 7 specific competencies within the team. focus on hiring good performers.

His 7 skills persona for the Unicorn AI team

Record the interpreter persona

The lifeblood of any AI project is data. Finding the right dataset, preparing the data, and ensuring continuous high quality are important skills. Because of the wealth of tribal knowledge about records, we need someone who specializes in tracking the meaning of data attributes and the provenance of various records. A related data challenge is managing multiple definitions of business metrics across an organization. In one of his projects of mine, I explored his eight definitions of “new customer every month” in sales, finance and marketing. A good starting point for this competency personality is being a traditional data warehouse engineer with strong data modeling skills and an innate curiosity about relating the meaning of data attributes to applications and business operations. .

Pipeline builder persona

A data pipeline is required to get data from multiple sources into an AI model. Within the pipeline, data is cleaned, prepared, transformed, and turned into ML functions. These data pipelines (known as extract, transform, load, or ETL in traditional data warehouses) can be very complex. Enterprises typically have pipeline jungles with thousands of pipelines built using disparate big data technologies such as Spark, Hive, and Presto. The Pipeline Builder persona focuses on building and operating pipelines at scale with appropriate robustness and performance. The best place to find this persona is a data engineer with years of experience developing batch and real-time event pipelines.

AI full-stack persona

AI is inherently iterative from design to training to deployment to retraining. Building an ML model requires hundreds of experiments with different combinations of code, features, datasets, and model configurations. This persona combines knowledge of the AI ​​domain with strong system building skills. They specialize in existing AI platforms such as Tensorflow, Pytorch, or cloud-based solutions such as AWS, Google and Azure. With the democratization of these AI platforms and the proliferation of online courses, this persona is no longer uncommon. In my experience, a solid background in software engineering and a curiosity to learn AI is a very effective combination. When hiring this talent, chances are you’ll come across geniuses who prefer to fly solo rather than team players. Be vigilant and eliminate early.

AI Algorithm Persona

Most AI projects rarely need to start from scratch or implement new algorithms. The role of this persona is to guide the team in the exploration area of ​​AI algorithms and technologies in the context of the problem. These help avoid dead ends during course correction and balance solution accuracy and complexity. Acquiring this talent is not easy given the high demand in places focused on AI algorithm innovation. If you can’t afford to hire a full-time person for this skill, consider hiring an expert as a consultant or advisor to his startup. Another option is to invest in training your full-stack team, giving them time to familiarize themselves with research progress and algorithm internals.

Data and AI operational personas

After an AI solution is deployed in production, it should be continuously monitored to ensure it is working properly. Many problems can occur during production. Data pipeline failures, poor data quality, under-provisioned model inference endpoints, inconsistent model prediction accuracy, uncoordinated changes to business metric definitions, and more. This persona focuses on building proper oversight and automation to ensure smooth operations. Compared to his traditional DevOps for software products, Data + AI Ops is very complex given the number of moving parts. Researchers at Google have correctly summarized this complexity as the CACE principle. A good starting point for finding this persona that changes everything is his DataOps Engineer, an experienced person who wants to learn about data and AI.

Hypothesis Maker Persona

AI projects are full of surprises. The path from raw data to actionable AI intelligence is not linear. You need flexible project planning, adapting based on confirming or rejecting assumptions about datasets, features, model accuracy, and customer experience. Traditional data analysts with experience working on multiple concurrent projects with tight deadlines are good candidates for discovering the personality of this ability. His talent for pursuing hypotheses and executing in parallel makes him an excellent project manager.

Personality of an influential owner

Impact owners are familiar with the details of how AI services are used to create value. For example, if this rep is using her AI to solve a problem related to improving customer retention, make sure that she fully understands her maps of customer acquisition, retention, and churn related journeys. will be You are responsible for defining how her team of specialists will support customer churn predictions from your AI solution and implement them to reduce churn. The best place to find this persona is within your existing business team. “Ideally, an engineer with strong product intuition and realism.” Without this persona, teams will only build what is technically feasible instead of thinking pragmatically about what they actually need in an end-to-end workflow to create value. .

In summary, these seven skill personas are must-haves for AI teams. The importance of these personas depends on the maturity of the data, the nature of the AI ​​problem, and the skills available across the broader data and application teams. For example, the role of the data interpreter is much more important in an organization with data in many small tables than in an organization with a few large tables. These factors should be considered when determining the appropriate seniority and cardinality for each skill persona within an AI team. Instead of waiting for the unicorn to show up, we hope you start building your AI team today.