Data Quality in the Agentic AI Era: Reliable Architectures, Java + AI, and Scalable Data Orchestration Kuldeep Singh hosts Abirami Sukumaran, Staff Developer Advocate at Google, who shares her two-decade journey from Infosys, IBM, and Citrix to Google, focused on data, analytics, Java, and databases. She highlights a key lesson from enterprise work: stand up for solutions you believe in while collaborating with stakeholders and aligning with user and business priorities. The discussion explores how data quality has become more critical in generative and agentic AI because systems influence decisions, while modern intelligent data platforms reduce the effort needed to achieve quality. Abirami explains enterprise challenges in connecting LLMs and agents to diverse, disparate data sources, emphasizing orchestration for performance, accuracy, and cost, plus security and privacy. She discusses Java’s evolving AI ecosystem, reliable agent architectures grounded in consolidated truth and semantic inference, and introduces Code Vipassana as an accessible hands-on learning initiative backed by Google.Time Stamps00:00 Highlights01:12 Show Welcome Intro01:28 Career Journey Highlights03:13 Conviction And Stakeholders07:58 Data Quality In GenAI10:44 Scaling LLM Data Stacks11:27 Orchestrating Diverse Data17:02 Java For Agentic Workflows21:56 Reliable Agent Architecture29:22 Code Vipassana Community34:50 Rapid Fire Round36:17 Wrap Up And ResourcesYouTube Channel: @techietalkswithkuldeep WhatsApp Channel: https://www.whatsapp.com/channel/0029Vb7Q4JHAjPXUt8hBgm2TLinkedIn: https://www.linkedin.com/company/techietalkswithkuldeepFacebook: https://www.facebook.com/profile.php?id=61574642927780Instagram: https://www.instagram.com/techietalkswithkuldeepFollow Abirami on LinkedIn: https://www.linkedin.com/in/abiramisukumaran/#ai #technology #machinelearning #database #datascience #softwareengineering #agenticai #techpodcast #techietalkswithkuldeep

SuperOwl COO Farah on Why Keywords Are Dead, Missed-Call Voice AI for India & Building Human-Centered TechIn this episode of Techie Talks with Kuldeep, host Kuldeep Singh interviews Mrs. Farah, co-founder and COO of SuperOwl, about her decade-plus entrepreneurial journey and the centrality of understanding customers. Farah recounts her 2008 work at Yahoo, including a provisional US patent around keyword trading, and contrasts keyword search then with today’s intent-driven systems that can choose and book services like a dentist appointment. She discusses her IISc research on intellectual property and early mobile payments, highlighting user trust barriers and technologies like NFC and USSD. Farah explains SuperOwl’s niche focus on missed calls for local businesses, the need for empathetic, low-latency, highly accurate voice AI, and India’s price sensitivity. She also shares her work on tactile children’s products like wooden blocks to balance gadget time, emphasizes empathy in tech, comments on women founders, and closes with lessons on ego, objections, and time management.Time Stamps00:00 Precaps01:24 Welcome And Guest Intro01:41 Entrepreneurship Lessons03:57 Yahoo Keywords To AI08:10 IISC Research And IP12:35 Engineering And Adoption17:11 Unit Economics In India20:10 Women In Deep Tech22:53 New Startup26:43 Human Empathy In AI29:31 Rapid Fire Round31:21 Closing ThoughtsYouTube Channel: @techietalkswithkuldeep WhatsApp Channel: https://www.whatsapp.com/channel/0029...LinkedIn: / techietalkswithkuldeep Facebook: https://www.facebook.com/profile.php?...Instagram: / techietalkswithkuldeep Follow Farah Ahmad on LinkedIn: / farah-ahmad-2a864939

Stop getting stuck in pilot projects. Learn how to scale your enterprise AI strategy beyond simple tests and into real value.Kuldeep Singh hosts Sandhya Gopalan, who shares many organizations treat AI like a science experiment, constantly stuck in the prototyping phase. We analyze why your machine learning operations are stalling and how to bridge the gap between initial tests and production-ready systems. It is not enough to have a working model if you cannot support it at scale.Moving past pilot projects requires a fundamental shift in how you analyze your cost, people, data, and technology. We discuss the necessary platform architecture to support sustainable growth and why Trusted AI is the missing piece for organizations trying to operationalize their efforts. If your team is running in place, this breakdown will help you reorient toward a strategy that actually works.Time stamps0:00 Highlights1:20 A Journey Through 21 Years of Tech5:48 Breaking Cultural and Architectural Barriers13:13 Avoiding Cloud Vendor Lock-in21:01 Operationalizing Trusted AI31:26 The Shift in DevOps Responsibilities38:19 Preventing Catastrophic AI Failures41:41 Closing remarksYouTube Channel: @techietalkswithkuldeep WhatsApp Channel: https://www.whatsapp.com/channel/0029...LinkedIn: / techietalkswithkuldeep Facebook: https://www.facebook.com/profile.php?...Instagram: / techietalkswithkuldeep Follow Sandhya Gopalan on LinkedIn: / sandhya-gopalan01

#ai #technology #machinelearning #mlops #softwareengineering #startup #techpodcast #techietalkswithkuldeep

Stop shipping features that don't move the needle. We are breaking down why your product execution might be the silent leak costing you revenue.High-growth companies often fall into the trap of confusing activity with progress. Just because your roadmap is full of shipped features does not mean your revenue generation is actually scaling. We analyze how to identify when your team is optimized for output rather than impact and why flat growth often points to a disconnect between engineering goals and business reality.Instead of blindly adding to the backlog, start asking your recent hires about how they navigate these complexities. Aligning your technical roadmap with actual market value is the only way to stop the bleed and get back to sustainable growth.Time Stamps0:00 Highlights1:08 Meet the Guest3:28 Shipping vs. Revenue: The Output Trap6:51 Identifying Friction in Your Development Process10:17 The Altitude Gap: Why Teams Fail to Align16:14 Playbooks for Engineering Velocity21:26 Finding True Levers for AI-Driven Growth23:41  Hard Truth & Advice about scaling product organization 24:46 Closing ThoughtsYouTube Channel: @techietalkswithkuldeep WhatsApp Channel: https://www.whatsapp.com/channel/0029Vb7Q4JHAjPXUt8hBgm2TLinkedIn: https://www.linkedin.com/company/techietalkswithkuldeepFacebook: https://www.facebook.com/profile.php?id=61574642927780Instagram: https://www.instagram.com/techietalkswithkuldeepFollow Ashvita MandalLinkedIn: https://www.linkedin.com/in/ashvita-m/#ai #technology #startups #developers #productmanagement #techleadership #softwareengineering #techpodcast #techietalkswithkuldeep

Scaling Fintech Systems & Tech Orgs: UPI Reliability, Monolith vs Microservices, and AI Blind Spots In this episode of Techie Talks with Kuldeep, host Kuldeep Singh speaks with industry veteran Kshitiz Agarwal about nearly two decades of experience across Samsung, Intel, OYO, Paytm, and ticketing/travel platforms. They discuss how architecture priorities shift in regulated, high-stakes systems like UPI, emphasizing fraud prevention, security, and reliability as critical to consumer trust. Kshitiz shares a framework for scaling engineering organizations by focusing on velocity (directional impact) over speed, using clear goals and iterative feedback loops. He explains why monolith-to-microservices migrations should be driven by specific problems and measurable outcomes, not trends, and outlines team-splitting approaches using product-aligned verticals and shared horizontal platforms with standard APIs. They also cover developing technical leaders through goals, autonomy, and minimal micromanagement, daily metric tracking for managers, career advice to learn the business and collaborate, and AI’s risk of encouraging careless attention to important details like sensitive logging.00:00 Precaps01:27 Introduction01:44 Career Journey Highlights03:13 Architecture For Trust05:19 Fintech Compliance Reality08:21 Scaling Teams With Velocity11:13 Monolith Vs Microservices14:34 Growing Technical Leaders16:37 AI Blind Spots For Leaders19:40 Rapid Fire Playbook22:25 Closing ThoughtsYouTube Channel: @techietalkswithkuldeep WhatsApp Channel: https://www.whatsapp.com/channel/0029Vb7Q4JHAjPXUt8hBgm2TLinkedIn: https://www.linkedin.com/company/techietalkswithkuldeepFacebook: https://www.facebook.com/profile.php?id=61574642927780Instagram: https://www.instagram.com/techietalkswithkuldeepFollow Kshitiz Agarwal on LinkedIn: https://www.linkedin.com/in/kshitiz-agarwal/#ai #technology #softwareengineering #systemdesign #techleadership #softwarearchitecture#cloudcomputing #techpodcast #techietalkswithkuldeepEngineering Leadership, System Design, Monolith to Microservices, Fintech Architecture, Techie Talks with Kuldeep, Scaling Engineering Teams, Software Architecture Best Practices

The Blackwell Paradox: GPU Utilization vs P99 Latency, AI Governance, and Edge AI Kuldeep Singh hosts Akhil Gupta to discuss operating AI systems in production, focusing on the “Blackwell Paradox” where pushing GPU utilization toward 95%+ via batching and speculative decoding can improve average latency but dangerously worsen P99, which is unacceptable for critical use cases like real-time fraud detection. Akhil argues CFO cost comparisons should weigh idle GPU headroom against outage and SLA-penalty costs, using an elevator-capacity analogy. They contrast classical ML for SLA-critical “fast lane” decisions with GenAI for “slow lane” post-transaction reasoning, behavioral analysis, and compliance support. In multi-tenant inference, Akhil proposes three-layer governance—admission control, per-tenant KV-cache isolation, and SLA-priority decoding—to prevent noisy workloads from impacting others. They also cover Edge AI with ExecuTorch, small/quantized models, and a routing metric favoring edge for latency/privacy and cloud for complex cross-domain needs, concluding governance must precede scaling AI systems.Understanding the Blackwell Paradox is critical for balancing GPU utilization and P99 latency in production AI systems.Kuldeep Singh hosts Akhil Gupta to break down why the Blackwell Paradox makes pushing GPU utilization above 95% often backfire. They explain how batching and speculative decoding might improve average latency but degrade P99 performance, creating unacceptable risks for real-time fraud detection.The discussion covers necessary AI governance frameworks, including admission control and priority decoding, to isolate noisy workloads. Akhil also shares a practical approach to inference optimization, comparing when to route traffic to edge AI versus the cloud, ensuring your infrastructure is built for reliability, not just raw compute density.Subscribe for weekly engineering and cloud architecture breakdowns. What is your biggest challenge when scaling LLM inference? Let us know in the comments.00:00 Highlights01:15 Introduction01:48 Akhil Career And AI Costs04:14 Blackwell Paradox Explained06:34 CFO View Idle Capacity09:40 Fast Lane Vs Slow Lane13:11 Shared Infra And Fallbacks16:06 Multi Tenant Governance Layers20:46 Edge AI Reality Check24:43 Route Requests Edge Or Cloud26:38 India Compute And Access28:08 Future Inference Ops Shifts30:58 Wrap Up And Takeaways32:43 Closing And Sign OffCONNECT WITH US 🌐YouTube Channel: @techietalkswithkuldeep WhatsApp Channel: https://www.whatsapp.com/channel/0029Vb7Q4JHAjPXUt8hBgm2TLinkedIn: https://www.linkedin.com/company/techietalkswithkuldeepFacebook: https://www.facebook.com/profile.php?id=61574642927780Instagram: https://www.instagram.com/techietalkswithkuldeepFollow Akhil Gupta on LinkedIn: https://www.linkedin.com/in/akkhil/👍 Like, subscribe, and comment your biggest challenge when scaling LLM Inference#ai #cloudcomputing #gpu #cloudsecurity #edgeai #aiintech #techpodcast #cloudcostoptimization #softwareengineering #cloudarchitecture #opensource #techietalkswithkuldeep

DevSecOps isn't just about tools — it's about culture. In this episode of Techie Talks with Kuldeep, host Kuldeep Singh sits down with Nishant Gupta, CEO & Co-Founder of DevSecOps.ai, to unpack how he scaled crypto exchange infrastructure, slashed DevOps SLAs, and saved companies over $1 million in cloud costs.We cover managing blockchain full nodes at scale, migrating stateful workloads to Kubernetes, cutting FinOps waste, replacing paid monitoring tools with open-source alternatives, building self-service infrastructure with security guardrails, and where AI actually fixes security gaps vs. where it's just hype.⏱️ TIMESTAMPS00:00 – DevSecOps is about culture, not just tools00:04 – Managing crypto exchange infra at high volume00:19 – Persistent storage: the first big challenge01:53 – Scaling ZebPay's blockchain nodes under compliance01:57 – Surviving state sync & consensus failures04:32 – Migrating full nodes from EC2 to EKS04:33 – Containerizing stateful blockchain workloads07:26 – FinOps: how $1M+ was saved in hybrid cloud09:23 – Fixing the architecture that was bleeding money13:08 – Replacing paid monitoring with open-source (Loki, Prometheus, Grafana)14:21 – Saving $25-30K/month on observability costs14:24 – Slashing DevOps SLA at WhiteHat Jr from 1 week to 2 days17:41 – Why engineering teams fail at automation19:16 – Self-service infra & security guardrails for developers20:50 – Why he left corporate after 18 years to found DevSecOps.ai24:04 – Where AI actually fixes security gaps (and where it's hype)30:14 – Rapid fire: legacy cloud practices that need to die31:52 – The psychological blind spot founders must fix32:56 – Life after leaving corporate: risk, finance & convictionCONNECT WITH US 🌐YouTube Channel: @techietalkswithkuldeep WhatsApp Channel: https://www.whatsapp.com/channel/0029Vb7Q4JHAjPXUt8hBgm2TLinkedIn: https://www.linkedin.com/company/techietalkswithkuldeepFacebook: https://www.facebook.com/profile.php?id=61574642927780Instagram: https://www.instagram.com/techietalkswithkuldeepFollow Nishant Gupta on LinkedIn: https://www.linkedin.com/in/nishant-gupta-83322210/👍 Like, subscribe, and comment your biggest DevOps/cloud challenge below!#DevSecOps #devops #fintech #edtech #cloudcomputing #kubernetes #aws #cloudsecurity #finops #blockchain #cryptoinfrastructure #aiintech #techpodcast #startupjourney #cloudcostoptimization #softwareengineering #cloudarchitecture #opensource #devopsengineering #techietalkswithkuldeep

In this episode of Techie Talks with Kuldeep, host Kuldeep Singh sits down with Mr. Daya Prakash, Founder of TalentOnLease and former CIO of LG India, to unpack why traditional corporate hiring is broken, how IT service companies can monetize bench resources without risking client confidentiality, and why "AI-powered" hiring platforms still need human judgment to earn trust. Daya also shares his entrepreneurial journey — moving from a corporate safety net to building TalentOnLease — and how it reshaped his perspective on risk and digital transformation.**Timestamps**0:00 – Highlights1:08 – Intro1:30 – Daya Prakash's career journey & how TalentOnLease started4:43 – Is traditional corporate recruitment fundamentally broken?8:14 – Monetizing IT bench resources without client confidentiality risk14:24 – Why recruitment agencies fail on large bulk-hiring demands18:30 – Deploying talent in 2-3 days: how quality is really ensured18:45 – "Everyone's overusing the word AI" — trust over hype24:30 – Short-term trials becoming long-term partnerships29:35 – From LG's CIO to founder: how his view on risk changed31:48 – Why entrepreneurs have no choice but to bounce back36:58 – Closing thoughts CONNECT WITH US 🌐YouTube Channel: @techietalkswithkuldeep WhatsApp Channel: https://www.whatsapp.com/channel/0029Vb7Q4JHAjPXUt8hBgm2TLinkedIn: https://www.linkedin.com/company/techietalkswithkuldeepFacebook: https://www.facebook.com/profile.php?id=61574642927780Instagram: https://www.instagram.com/techietalkswithkuldeepFollow Daya Prakash on LinkedIn: https://www.linkedin.com/in/dayaprakash/#ai #technlogy #futureofwork #techhiring #techcareer #startupjourney #hrtech #entrepreneurship #leadership #techpodcast #mindset #techietalkswithkuldeep