The Tech Trek

Stop treating data governance as a "data cop" function and start using it as a high ROI offensive weapon. In this episode, Peter Kapur, Head of Data Governance and Data Quality at CarMax, breaks down how to move beyond defensive compliance to drive profitability, customer experience, and better data science outcomes.Critical Insights for LeadersShift from defense to offense Data defense covers the mandatory regulatory and legal requirements like privacy and cybersecurity. Data offense involves everything else that hits your bottom line, such as investing in data quality to save or make money.Prioritize problems over frameworks Avoid bringing rigid policies and "data geek" terminology to business leaders. Instead, spend time listening to their specific data struggles and apply governance capabilities as solutions to those problems.Data quality makes governance tangible Without high quality data, governance is just a collection of abstract policies. Improving data quality empowers data scientists to produce better models and gives analytics teams the ability to discover and trust their data.Key Moments in the Conversation02:41 Defining the clear line between defensive regulation and offensive growth 06:03 Why data quality and data governance must sit together to be effective 11:00 Shifting from "data school" to "business school" to communicate value 13:12 Quantifying the ROI of data governance through customer wins and time savings 18:35 Actionable advice for starting an offensive strategy from scratch Wisdom from the Episode"If we meet the laws, we meet the regulations, we meet the legal, how do we leverage our data? It is a mindset shift versus, let me lock my data down, no one use it." Tactical Advice for ImplementationEnsure adoption through personalization Design tools and processes that are personalized to specific roles so they feel like a natural part of the workflow rather than a burden.Focus on the eye of the consumer Treat every person in the organization as a "data citizen" and remember that data quality is ultimately defined by the needs of the people consuming it.Join the ConversationSubscribe to the podcast on your favorite platform to catch every episode. Follow us on LinkedIn to stay updated on the latest trends in data leadership.

Afrooz Ansaripour, Director of Data Science at Walmart, joins the show to explain how global leaders are shifting from simple historical tracking to predicting psychological triggers and customer intent. This episode explores the evolution of customer intelligence and how Generative AI is turning massive data sets into personalized, value driven experiences. Listeners will learn how to balance hyper personalization with foundational privacy to build lasting consumer trust.Key InsightsPredict intent rather than just reporting past transactions to understand why a customer is with the brand.Use Generative AI as an explainability layer to transform complex data platforms from black boxes into conversational tools.Prioritize customer trust as a critical part of the user experience rather than just a legal requirement.Integrate digital and physical signals to create a 360 degree view that reveals insights which would otherwise be invisible.Focus on rapid technology adoption and curiosity as the primary drivers of success in modern AI teams.Timestamped Highlights01:51 Identifying the challenges and opportunities when managing millions of real time signals.06:43 Strategies for showing genuine value to the customer without making them feel like just a part of a sale.09:51 How LLMs are fundamentally changing the way data teams interpret unstructured feedback and behavioral patterns.14:42 Managing privacy and ethical data practices while building personalized conversational AI.19:14 Stitching together the online and offline journey to create a seamless customer experience.22:52 The necessary evolution of data science skills toward storytelling and execution bias.A Powerful Thought"Personalization should never come at the expense of customer trust." Tactical StepsCombat the garbage in garbage out problem by refining cleaning processes to handle modern AI requirements.Build an interactive layer or chatbot on top of data products to make insights instantly accessible and automated.Translate technical insights into real world decisions to ensure customers actually benefit from data models.Next StepsSubscribe to the show for more insights into the future of tech. Share this episode with a peer who is currently navigating the complexities of customer data.

Shahryar Qadri, CTO of OneImaging, joins me to unpack a hard truth about healthcare tech: the goal is not to remove humans, it is to give them more room to be human.We talk about where cost “optimization” actually helps patients, why radiology is a perfect fit for AI but still held back by data access, and how better workflows can improve trust, speed, and outcomes without losing the human touch.OneImaging sits in the radiology benefits space, helping members book imaging in a national network with more transparency and a high touch booking experience, while helping employers cut imaging costs significantly.Key takeaways• The “human touch” in healthcare is not going away, the better play is using tech to increase capacity so caregivers can spend more time being caregivers• Cost optimization is not always about paying less for expertise, it is often about wasting less human time, improving trust, and removing friction around services• Healthcare still runs on outdated plumbing in places you would not expect, including fax based workflows that slow everything down• Radiology is one of the best real world use cases for AI, but the bigger blocker is getting access to imaging data in usable form, not model capability• Your health data is already “there”, but it is not working for you yet. The next wave is tools that scan your longitudinal record and surface what to ask your doctor about, so you can be a stronger advocate for your own careTimestamped highlights• 00:36 What OneImaging actually does, and why “transparent imaging” is more than a pricing story• 02:00 Why healthcare stays personal, and how tech should increase capacity instead of replacing care• 03:36 The real definition of cost optimization, commodity versus service, and where trust matters• 07:01 The surprising reality of imaging ops, why it still feels like 1998, and what gets digitized next• 17:19 AI in radiology is real, but the data access and interoperability gap is the bottleneck• 24:21 Your CDs are full of value, the problem is we do almost nothing with that data todayA line worth replaying“These LLM models are the worst that they’ll ever be today. They’re only going to get better and better and better.”Call to actionIf this episode sparked a new way of thinking about healthcare tech, follow The Tech Trek on your podcast app, share it with a friend in product or engineering, and connect with me on LinkedIn for more conversations like this.

Swarupa Mahambrey, Vice President of Software Engineering at The College Board, breaks down what tech debt really looks like in a mission critical environment, and how an engineering mindset can prevent it from quietly choking delivery. She shares a practical operating model for paying down debt without stopping the roadmap, and the cultural habits that make it stick.You will hear how College Board carved out durable space for engineering excellence, how they use testing and automation to protect reliability at scale, and how to make the trade offs between features, simplicity, and user experience without slowing the team to a crawl.Key Takeaways• Tech debt behaves like financial debt, delay the payment and the interest compounds until even simple changes become painful• A permanent allocation of capacity can work, dedicating 20 percent of every sprint to tech debt can reduce support load and improve delivery• Shipping more features can slow you down, simplifying workflows and validating with real usage can increase velocity and reduce tickets• Resilience is not about avoiding every failure, it is about designing for graceful degradation so spikes and outages become small blips instead of crises• Automation is not “extra,” it is part of the definition of done, including unit tests as acceptance criteria and clear code coverage expectationsTimestamped Highlights• 00:00 Why tech debt is a mindset problem, not just a backlog problem• 01:00 Tech debt explained with a real example, what happens when a proof of concept becomes production• 03:45 The feature trap, how “powerful” workflows can overwhelm users and explode maintenance costs• 11:03 Engineering Tuesday, one day a week to strengthen foundations, not ship features• 14:39 Stability vs resilience, designing systems that bend instead of shatter• 20:06 Testing and automation at scale, unit tests as a requirement and code coverage guardrailsA line worth keeping“If we don’t intentionally carve out space for engineering excellence, the urgent will always crowd out the important.”Practical moves you can steal• Protect a fixed slice of capacity for tech debt, make it part of the operating model, not a one time cleanup• Treat automation as acceptance criteria, no test, no merge, no release• Use pilots and targeted releases to learn early, then iterate based on metrics and real user behavior• Design for graceful degradation with retries, fallback paths, and clear failure visibilityCall to actionIf this episode helped you think differently about tech debt and engineering culture, follow The Tech Trek, leave a quick rating, and share it with one engineer who is fighting fires right now.

Software is still eating the world, and AI is speeding up the clock. In this episode, Amir talks with Tariq Shaukat, co CEO at Sonar, about what it really takes for non tech companies to build like software companies, without breaking trust, security, or quality. Tariq shares how leaders can treat AI like a serious capability, not a shiny add on, and why clean code, governance, and smart pricing models are becoming board level topics. Key Takeaways• “Every company is a software company” does not mean selling SaaS, it means software is now core to differentiation, even in legacy industries. • The hardest shift is not tools, it is mindset: moving from slow, capital style planning to fast iteration, test, learn, and ship. • AI works best when leaders stay educated and involved, outsourcing the whole strategy is a real risk. • “Trust but verify” needs to be a default posture, especially for code generation, security, and compliance. • Pricing will keep moving toward value aligned consumption models, not simple per seat formulas. Timestamped Highlights• 00:56 What Sonar does, and why clean code is really about security, reliability, and maintainability • 05:36 The Tesla lesson: mechanics commoditize, software becomes the experience people buy • 09:11 Culture plus education: why software capability cannot live in one silo • 14:21 Cutting through AI hype with program discipline and a “trust but verify” mindset • 18:23 Boards, governance, and setting an “acceptable use” policy for AI before something goes wrong • 25:18 How software pricing changes in an AI world, and why Sonar prices by lines of code analyzed A line worth saving:“Define acceptable risk as opposed to no risk.” Pro Tips you can steal• Write down what you want AI to achieve, the steps to get there, and the metric you will use to verify outcomes. • For code generation, scan and review before shipping, treat AI output like a draft, not a final answer.• Set clear rules for what is allowed with AI inside the company, then iterate as you learn. Call to ActionIf you want more conversations like this on software leadership, AI governance, and building real impact, follow The Tech Trek and subscribe on your favorite podcast app. If someone on your team is wrestling with AI rollout or developer productivity, share this episode with them.