Troy Redding | Director of AI Strategy & Adoption work sample
Expand useful AI across Hazen.
Hazen has already demonstrated practical AI applications in water treatment and operations. I would build on that foundation by identifying where AI can deliver further value, supporting adoption across teams, and measuring results.
Three first priorities
These are initial assessments.
01
Understand the current work
Assess existing capabilities, workflow friction and access readiness before choosing a solution.
02
Choose a small portfolio
Compare candidate uses by value, readiness and risk, then agree on scope and ownership.
03
Earn expansion with evidence
Evaluate quality, adoption, controls and full cost before continuing, fixing, stopping or expanding.
Experience I would bring
Team leadership and technical delivery
PolicySight Founder and Principal
As Founder and Principal of PolicySight, I hired and developed an 11 person multidisciplinary team of two computer scientists, three MBA team members and six lobbyists. I lead technical delivery and resource allocation across client engagements, connecting research, analytics and AI with executive decisions.
Adoption and operational AI
Applied AI at Metron
At Metron, I progressed from Data Analyst to Data Engineer to Applied AI Engineer reporting to the CTO. I built companywide AI and Power BI for schedules, materials, costs, risk and quality, led adoption and training across functions, and led the Data Task Force.
Evaluation and responsible controls
Production ML and grounded AI
My production ML work used leakage aware time splits, entity holdouts, explainability and human review. At PolicySight, I built client applications, data pipelines and multistep AI agents with source grounding, retrieval evaluation and hallucination checks, helping teams rely on inspectable evidence.
People leadership and executive counsel
Direct management at Baker Hughes
At Baker Hughes, I managed four direct reports and later four to six as a senior manager. I coordinated technical, legal and commercial work, advised regional presidents, business unit vice presidents and general counsel, and directed APCO Worldwide and other external specialists.
Hazen Intelligence
Make project history, standards, lessons and internal expertise easier to find and reuse.
Flagship candidate
I would assess existing tools first, then test a secure knowledge layer with clear source links and permission controls. The worked example covers supported answers, restricted access and insufficient evidence using fictional records.
Explore the worked exampleRelevant experience
At PolicySight, I built RAG workflows with source grounding, retrieval evaluation, hallucination checks and human review.
The first 90 days
Days 1 to 30
Establish the baseline
Review existing tools, workflows, sources and access readiness with the people who own the work.
Days 31 to 60
Agree on scoped pilots
Choose no more than three efforts with named owners, qualified reviewers and evaluation criteria.
Days 61 to 90
Review the evidence
Use early findings to continue, fix, stop or prepare an expansion decision. Do not assume every pilot should proceed.
A useful starting conversation
Which existing capabilities and workflows should shape the first 90 days?
I would welcome a discussion about the priorities that best fit Hazen's work and how I would approach delivery.
Troy Redding | Candidate for Director of AI Strategy & Adoption