“How Pavel Spesivtsev Argues That Knowledge Infrastructure Matters More Than AI Models”

“How Gautam Choudhury Built Healthcare AI That Prioritizes Interoperability, Reliability, and Trust at Scale”

EPISODE SUMMARY: In this episode of Engineering Choices You Have to Defend, host Nicola Onassis sits down with Gautam Choudhury, Co-Founder and CTO of Zyex AI, to discuss one of healthcare AI's most difficult engineering challenges: building intelligent systems that work reliably across fragmented healthcare environments. Rather than optimizing for a single electronic medical record (EMR) platform, Gautam and his team made the difficult architectural decision to build an EMR-agnostic platform from day one. Serving healthcare organizations that may operate hundreds or even thousands of EMR instances, Zyex AI focuses on automating care coordination, scheduling, outreach, documentation, and operational workflows across highly fragmented systems. The conversation explores why interoperability should be treated as a reliability problem instead of simply an API integration challenge. Gautam explains how healthcare workflows extend far beyond structured APIs, requiring intelligent automation through robotic process automation (RPA), adaptive AI agents, and resilient workflow orchestration capable of handling real-world operational complexity. A major focus of the discussion is the balance between AI automation and human oversight. Rather than replacing healthcare professionals, Zyex AI uses confidence thresholds, governance, and human checkpoints to ensure sensitive clinical and operational decisions remain accountable while AI eliminates repetitive administrative work. For engineering leaders building AI systems in regulated industries, this episode offers valuable lessons on designing deployable architectures, building trust into AI systems, and creating operationally resilient platforms that succeed in production—not just in demonstrations. KEY TAKEAWAYS: * Interoperability should be treated as an operational reliability problem, not simply an API integration project * Building EMR-agnostic architecture creates long-term scalability across fragmented healthcare environments * Healthcare AI must integrate with multiple systems beyond EMRs, including telephony, fax, scheduling, and manual workflows * AI-powered RPA creates more resilient automation by adapting to changing interfaces and operational variability * Human oversight remains essential for clinical ambiguity, regulatory accountability, and low-confidence AI decisions * Infrastructure flexibility is critical for healthcare organizations with varying compliance and deployment requirements * Deployable architecture often matters more than model sophistication in healthcare AI * Trust, governance, and operational reliability drive adoption more than raw AI performance * Engineering teams should optimize for production reliability rather than polished demonstrations * Successful healthcare AI platforms are built to survive operational complexity at scale CONNECT WITH GAUTAM CHOUDHURY: LinkedIn: linkedin.com/in/gautamchoudhury2007 [ linkedin.com/in/gautamchoudhury2007] Website: zyex.ai [zyex.ai] Listen Now & Subscribe: Apple Podcasts, Spotify, Amazon Music, YouTube, iHeartRadio, Captivate, or wherever you get your podcasts. "Engineering Choices You Have to Defend explores the real technical decisions behind AI systems, enterprise architecture, and scalable software engineering.

“How Eban Bisong Transformed Engineers into AI Orchestrators to Eliminate Delivery Bottlenecks”

Episode Summary: In this episode of Engineering Choices You Have to Defend, host Nicola Onassis sits down with Eban Bisong to discuss how AI-native workflows are reshaping software engineering teams and changing what it means to be an engineer. While leading engineering at Part DNA, Eban faced a challenge familiar to many growing organizations: a small team supporting multiple clients, constant context switching, increasing delivery demands, and pressure to maintain quality while moving faster. Traditional approaches were no longer enough. Rather than simply introducing AI coding tools, Eban led a broader organizational transformation that redefined how work moved through the company. By integrating AI agents into engineering, support, documentation, ticket creation, code review, testing, and knowledge management workflows, the team dramatically reduced operational bottlenecks and increased delivery capacity without increasing headcount. A key part of the transformation was the introduction of an AI teammate named R2-D2, powered by OpenClaw. Initially deployed as a refactoring and code-quality agent, R2-D2 evolved into a company-wide knowledge assistant capable of supporting engineering, customer support, documentation, and operational workflows. The result was a system where AI handled repetitive execution tasks while humans focused on judgment, architecture, customer conversations, and product strategy. The conversation explores how engineering roles are evolving from writing code to orchestrating systems that generate code, why specification quality is becoming more important than technical implementation, and how organizations can build AI-native processes that improve both speed and quality. For engineering leaders, this episode offers a practical framework for moving beyond AI experimentation and building organizations where humans and agents work together to create scalable, high-performing engineering systems. Key Takeaways: • AI adoption requires a mindset shift, not just new tools • Engineers are increasingly becoming orchestrators rather than code producers • AI agents can eliminate context-switching bottlenecks across organizations • Knowledge management and specifications are critical for successful AI workflows • Support, documentation, and engineering processes can all benefit from AI automation • Verification systems must scale alongside development velocity • AI agents should be separated across testing and implementation workflows • Product thinking and systems thinking are becoming more valuable than framework expertise • Organizations should optimize for judgment and decision-making, not manual execution • Successful AI-native teams focus on improving systems rather than fixing isolated problems Connect with Eban Bisong: * LinkedIn: linkedin.com/in/ebanbisong [linkedin.com/in/ebanbisong] * Website: ebanbisong.com [ ebanbisong.com] Listen Now & Subscribe: Apple Podcasts, Spotify, Amazon Music, YouTube, iHeartRadio, Captivate, or wherever you get your podcasts. "Engineering Choices You Have to Defend explores the real technical decisions behind AI systems, enterprise architecture, and scalable software engineering.

“How Pavel Spesivtsev Argues That Knowledge Infrastructure Matters More Than AI Models”

Episode Summary: In this episode of Engineering Choices You Have to Defend, host Nicola Onassis sits down with Pavel Spesivtsev, CTO, AI strategist, and agentic engineering practitioner, to explore why many AI-driven software initiatives fail long before coding becomes the problem. After spending the last eighteen months helping organizations implement agentic development workflows, Pavel has observed a surprising pattern: the models themselves are rarely the weakest link. Instead, failures typically emerge from incomplete specifications, missing organizational knowledge, weak governance, and poor context management. Pavel explains why traditional software development assumptions are being challenged by agentic engineering. While Agile methodologies were designed around human decision-making and implementation, AI agents require far more structured specifications and complete knowledge systems to operate effectively. When requirements contain gaps, agents fill them with assumptions drawn from training data, often leading to unexpected or incorrect outcomes. The conversation explores Pavel’s concept of “Gap Trap,” a framework designed to identify missing requirements before they enter an agentic workflow. He also discusses why knowledge bases and ontologies are becoming critical infrastructure for AI-powered development, how retrieval systems can introduce hidden hallucination risks, and why context engineering is rapidly becoming one of the most valuable skills in modern software organizations. Pavel shares his perspective on the evolution of software engineering roles as AI adoption accelerates. As implementation becomes increasingly automated, engineers are spending less time writing code and more time designing systems, orchestrating agents, validating outputs, and building the knowledge frameworks that guide intelligent systems toward reliable outcomes. For engineering leaders, this episode highlights a major shift in software delivery: as coding becomes increasingly automated, competitive advantage will come from designing better systems, creating higher-quality specifications, and building the knowledge infrastructure that enables AI agents to make reliable decisions. Key Takeaways: • Most agentic AI project failures stem from specification and knowledge gaps, not model quality • Incomplete requirements cause AI agents to make unpredictable assumptions • Knowledge bases and ontologies are becoming critical infrastructure for AI systems • Context engineering is emerging as a core engineering discipline • Retrieval systems can introduce hidden hallucination risks when information is incomplete • Software engineers are evolving from code authors into system architects and orchestrators • Agentic workflows require stronger specification practices than traditional Agile processes • Documentation is increasingly becoming operational infrastructure, not just reference material • Governance, security, and knowledge management are essential for successful AI adoption • Organizations should focus on knowledge quality before investing heavily in AI tooling Connect with Pavel Spesivtsev: * LinkedIn: linkedin.com/in/pspesivt [inkedin.com/in/pspesivt] Listen Now & Subscribe: Apple Podcasts, Spotify, Amazon Music, YouTube, iHeartRadio, Captivate, or wherever you get your podcasts. "Engineering Choices You Have to Defend explores the real technical decisions behind AI systems, enterprise architecture, and scalable software engineering.

"How Paul Baker Stopped Feature Development to Save Engineering Velocity"

Episode Summary: In this episode of Engineering Choices You Have to Defend, host Nicola Onassis sits down with Paul Baker to discuss one of the most difficult decisions an engineering leader can make: stopping feature development in order to rebuild the engineering foundation. While working at Capshare, Paul inherited a growing product with strong market traction but a fragile engineering system plagued by regressions, manual testing, multi-day deployments, and the absence of automated quality controls. Faced with mounting production issues and increasing customer risk, Paul proposed an unconventional solution: pause all new feature development for an entire quarter and focus exclusively on improving software quality, testing, and deployment infrastructure. Paul shares how the team implemented automated testing, continuous integration, and systematic refactoring strategies to transform a legacy codebase into a maintainable platform capable of supporting future growth. He explains why engineering foundations are often the true drivers of delivery velocity and how technical debt can quietly undermine even successful products. The conversation also explores the evolving role of AI in software development, including the use of LLMs to accelerate legacy system modernization, generate large-scale test suites, and support engineering workflows. Paul offers practical insights into the limitations of agentic coding systems, the importance of prompt accuracy, and why human oversight remains essential as AI-assisted development becomes more common. For engineering leaders, this episode provides a powerful reminder that sustainable innovation depends on confidence in deployment, disciplined engineering practices, and investing in the foundations that make rapid delivery possible. Key Takeaways: • Engineering velocity depends on strong testing and deployment foundations • Pausing feature development can sometimes accelerate long-term delivery • Automated testing reduces production regressions and deployment risk • Legacy systems can be modernized through incremental refactoring strategies • Continuous integration creates confidence in software changes • Golden master testing can help stabilize complex legacy applications • AI can dramatically accelerate test generation and modernization efforts • Agentic coding systems still require human guidance and oversight • Deployment anxiety often reveals gaps in engineering infrastructure • Successful engineering organizations continuously invest in foundational quality Connect with Paul Baker: * LinkedIn: linkedin.com/in/pbaker3 [linkedin.com/in/pbaker3] * Website: paulbaker3.com [paulbaker3.com] Listen Now & Subscribe: Apple Podcasts, Spotify, Amazon Music, YouTube, iHeartRadio, Captivate, or wherever you get your podcasts. "Engineering Choices You Have to Defend explores the real technical decisions behind AI systems, enterprise architecture, and scalable software engineering.

“How Alexander Smirnoff Built Practical Enterprise AI Systems by Combining GenAI with Traditional NLP”

Episode Summary: In this episode of Engineering Choices You Have to Defend, host Nicola Onassis sits down with Alex Smirnoff to explore how enterprise AI systems can deliver real business value without replacing the proven infrastructure that already works. At Luminoso, Alex has spent years building large-scale NLP and text analytics systems that help enterprises analyze customer reviews, semantic search data, and large document collections. When generative AI rapidly entered the market, the company faced pressure from customers and stakeholders to “AI everything” overnight. Instead of rebuilding the platform around large language models, Luminoso chose a hybrid architecture that combined traditional NLP algorithms, semantic search, classification systems, and retrieval pipelines with modern GenAI reasoning capabilities. Alex explains why many older NLP tools still outperform LLMs for specific tasks like classification and keyword extraction, and how GenAI works best as an intelligent reasoning layer on top of existing systems. The conversation also explores hallucinations in enterprise environments, RAG pipeline design, grounding responses in source data, and the growing gap between flashy AI demos and production-ready enterprise systems. For engineering leaders, this episode highlights an important lesson: practical AI systems are rarely built by replacing everything — they succeed by combining proven infrastructure with new reasoning capabilities in thoughtful, cost-effective ways. Key Takeaways: * Traditional NLP tools still outperform LLMs for many specialized tasks * GenAI works best as a reasoning layer on top of existing systems * Hybrid AI architectures reduce cost and improve scalability * Enterprise AI systems must ground responses in customer data * RAG pipelines require careful tuning and retrieval quality management * Hallucination control is critical in business environments * There is a major gap between AI demos and production systems * Replacing entire platforms with GenAI often creates unnecessary complexity * Engineering teams should focus on business use cases, not AI hype * Successful AI adoption requires experienced implementation and planning Connect with Alex Smirnoff: * LinkedIn: Alex Smirnoff — linkedin.com/in/alex-smirnoff-34a13135 [http://linkedin.com/in/alex-smirnoff-34a13135] Listen Now & Subscribe: Apple Podcasts, Spotify, Amazon Music, YouTube, iHeartRadio, Captivate, or wherever you get your podcasts. "Engineering Choices You Have to Defend explores the real technical decisions behind AI systems, enterprise architecture, and scalable software engineering.