Techie Talks with Kuldeep

Techie Talks with Kuldeep

🌟Whether you're a student exploring tech, a professional leveling up your career, or a startup founder navigating innovation, this is your space to grow. 🔥What you’ll discover here: - Learn from the best - Stay ahead in tech - Fuel your ambitions - Honest discussions about challenges, failures, mindset, growth, and learning in the tech world - Practical advice for students, job seekers, tech enthusiasts, and startup builders - Inspiring career journeys of successful tech professionals - Deep dives into cutting-edge topics Artificial Intelligence, Cloud Computing, Cybersecurity, Blockchain

  1. 6 days ago

    Is Your AI Infrastructure Ready? GPU Utilization , Governance, Edge AI and The Blackwell Paradox

    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

  2. 30 Aug

    DevSecOps in Fintech & EdTech: Crypto Infra, AI & Million-Dollar Cost Cuts |Co-Founder Nishant Gupta

    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

  3. 23 Aug

    Is Your Recruitment Strategy Failing? The Truth About Bulk Hiring | AI Hiring reality check

    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

  4. 16 Aug

    Edge AI Strategy: Why You Need to Pivot |Real Reason EdgeAI Is Taking Over

    In this Techie Talks with Kuldeep episode, Kuldeep Singh interviews Kartikeya, a senior developer advocate and senior engineer at Qualcomm, about deploying and optimizing AI models on edge devices. They discuss why many web and cloud engineers struggle with edge deployment due to limited understanding of ML fundamentals, inference engines, and model optimization techniques like quantization and pruning. Kartikeya explains the TensorFlow Lite to LiteRT transition as primarily a front-end change, with Qualcomm’s QNN runtime and NPUs enabling private, offline on-device inference, and highlights Qualcomm AI Hub for model conversion, profiling, and deployment across Snapdragon-powered devices. They cover quantization trade-offs (PTQ vs QAT), data quality impacts on accuracy, kernel/runtime considerations for power and performance, the feasibility of local RAG with vector databases, and why enterprises adopt edge AI for cost reduction, privacy, and hybrid edge-cloud architectures.TimeStamps00:00 Precaps01:15 Meet the Guest02:38 Misconceptions and Model Shrinking05:50 LiteRT and Snapdragon Runtime09:45 Why Edge AI Matters12:06 Local LLM IDE Demo14:13 Quantization Beyond PTQ19:00 Battery and Kernel Optimization23:18 Local RAG on Device25:29 Enterprise Hybrid Edge Cloud30:59 Edge AI Learning Path35:47 Edge AI Is Everywhere38:41 Wrap Up and GoodbyeCONNECT 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 Kartikey Rawat on LinkedIn:https://www.linkedin.com/in/carrycooldude/#ai #technology #edgeai #edgecomputing #machinelearning #artificialintelligence #aiedge #quantization #modeloptimization #tinyml #hybridai #qualcomm #deeplearning #futureofai #techpodcast

  5. 16 Aug

    The FDE Engineer: A Game Changer in AI | Forward Deployed Engineer | ft. Vikas Sinha

    What is a Forward Deployed Engineer (FDE)? Enterprise GenAI, Data Governance & Agentic AI in Production | A Must-Have Role in AIIn this episode of “Techie Talks with Kuldeep,” host Kuldeep Singh interviews Vikas Sinha, an AI consulting manager with 14 years of experience, about the rising demand for Forward Deployed Engineers (FDEs) in enterprises. Vikas explains that FDEs combine the roles of AI engineer, consultant, and domain expert, with end-to-end ownership of business outcomes and KPIs rather than only model metrics. He argues that most enterprise AI challenges are data problems, emphasizing data lineage, harmonization, governance, security, and transparency to move solutions from POC to production. The conversation covers agentic AI complexities across data, orchestration, tools, evaluations, monitoring, and feedback loops, and highlights key skills for aspiring FDEs: continuous learning, a T-shaped profile, strong communication, scalable architecture principles, and evaluation-first development.Time Stamps00:00 Highlights01:13 Introduction02:33 What Is An FDE04:04 Ownership And Business KPIs05:12 Why Enterprises Need FDEs09:01 Data Governance First12:54 Agentic AI Production Gaps18:34 Skills To Become FDE24:11 Advice For New Developers27:45 Rapid Fire Essentials29:58 Wrap Up And TakeawaysCONNECT 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 Vikas Sinha on LinkedIn:https://www.linkedin.com/in/vikassinha167/#ai #technology #genai #techpodcast #engineer #engineering #consultant #techskills #techcareers #futureofwork #techietalkswithkuldeep

  6. 26 Jul

    Your AI Agents Are Running Wild. Who’s Actually Controlling Them? | Jim Bugwadia, Co-Founder & CEO

    In this episode of Techie Talks with Kuldeep, host Kuldeep Singh sits down with Jim Bugwadia, Co-founder & CEO of Nirmata, Inc to explore AI governance, policy as code, and how Kyverno is evolving from Kubernetes policy enforcement to full-stack AI controls.Timestamps:0:00 – Teaser highlights1:19 – Meet the Guest1:39 – Jim's background: From Mumbai to Silicon Valley & telecom to cloud native2:49 – Staying hands-on as a founder3:19 – Why decoupling policy from applications matters in the age of AI4:03 – The OODA Loop: How decision-making frameworks apply to Kubernetes governance5:20 – Policy as code, Kyverno, and the enforcement + decision-making two-layer model6:38 – Why Nirmata, Inc chose Kubernetes CRDs (not DSLs) for Kyverno9:13 – CEL (Common Expressions Language) adoption in Kyverno for better performance12:44 – How Kyverno handles webhook latency, high throughput & large policy sets17:02 – Open source strength: Community, KubeCon & policy best practices17:44 – Nirmata's AI Controls: Governing LLM API calls, MCP requests & agent-to-agent calls18:31 – Risks in enterprise AI: Privacy, data leakage, token costs & productivity20:22 – Governance for AI engineering (LLMs as the new database layer)21:11 – Zero trust for AI agents: Covering all layers with Kyverno22:22 – AI governance latency: Why it's microseconds, not seconds24:04 – AI Controls as middleware: Caching, routing & lowering token costs26:27 – Balancing security guardrails without blocking developer productivity28:11 – Self-service exception management for developers30:25 – Security culture: Collaboration between dev and security teams31:00 – Rapid Fire Round31:32 – Language pick for next 5 years: Go/Golang32:44 – Top Kyverno configuration mistakes in production34:07 – Emerging CNCF projects to watch: Agent Substrate36:19 – Closing thoughts & where to try Kyverno and AI ControlsCONNECT 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 Jim Bugwadia on LinkedIn:https://www.linkedin.com/in/jimbugwadia/🔗 Learn more about Nirmata: https://nirmata.com🔗 Try Kyverno: https://kyverno.io#ai #technology #futureofwork #techpodcast #kubernetes #devops #llms #machinelearning #techietalkswithkuldeepkubernetes, kyverno, policyascode, aigovernance, devops, cloudnative, aicontrols, techietalkswithkuldeep, platformengineering, devsecops , llm , aiagents , mlops, kubernetessecurity, openSource , golangprogramming , infrastructureascode

  7. 11 Jul

    UPSC Aspirant to DevOps Engineer | Akshay Kumar's Tip for Landing a Job in Industry

    From UPSC to DevOps: Akshay’s Cloud-Native Journey, Kubernetes Roadmap & Using AI the Right Way. In this episode of Techie Talks with Kuldeep, Akshay Kumar shares how he transitioned from five UPSC IAS attempts and a mechanical engineering background into tech, eventually becoming a project manager at Cognite.ai after roles including NIC and open-source evangelism at Obmondo. He explains how COVID and a final UPSC attempt led to a career switch, supported by Vipassana meditation, self-learning, an unpaid Linux-focused internship, and hands-on DevOps skills that helped him clear the NIC interview. Akshay highlights that certifications aren’t enough, emphasizing community learning through Twitter/LinkedIn, events, mentoring, and building the Growin community. He outlines core pillars for DevOps/platform engineering—Linux/networking, Git/GitHub, Docker, Kubernetes, IaC (Terraform), and cloud platforms—and advises beginners to learn Kubernetes through basics and projects. He also shares how to use LLMs for troubleshooting via step-by-step prompting without copy-pasting, and discusses AI-related entry-level job anxiety, stressing continuous learning and adaptability.00:00 Highlights01:08 Meet the Host and Guest02:43 The Switch Decision04:26 Choosing Cloud and DevOps06:00 Breaking into NIC08:49 Beyond Certifications12:31 Community and Open Source16:17 DevOps Roadmap Basics18:52 Learning Kubernetes Right21:24 Using AI to Learn24:53 AI Troubleshooting Workflow27:19 AI Anxiety and Staying Hireable30:26 Advice for Non Tech Learners32:06 Closing-------------------------------------------------------------------------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 Akshay on LinkedIn: https://www.linkedin.com/in/akshayktwt/#devops #kubernetes #cloudnative #techietalkswithkuldeep #techpodcast #technology #futureofwork #ai #careergrowth #professionaldevelopment

  8. 4 Jul

    The Unstoppable Mindset That Scales Entire Organizations | ft. Sanjay Singh

    In this masterclass episode of Techie Talks with Kuldeep, host Kuldeep Singh sits down with industry veteran Sanjay Singh (Fractional CTO and Angel Investor). Sanjay shares raw, un-sugarcoated insights from his decades-long journey building and scaling tech—from the pre-TCP/IP days of telecom to handling catastrophic scale as a multi-time CTO at Paytm, managing 600+ engineers at Tiket.com, and founding an acquired healthcare analytics startup.Sanjay breaks down why most modern software engineers fail at core engineering fundamentals, the truth about organizational design vs. throwing processes at low talent density, and the exact playbook he uses to evaluate startups as an angel investor looking at hundreds of deals a month.Whether you are a software engineer, a tech leader navigating cross-border teams, or an aspiring founder, this episode provides a realistic blueprint for balancing technical elegance with business viability.⏱️ CHAPTERS (TIMESTAMPS)00:00 - Key Highlights01:39 - Introduction: Meet Sanjay Singh (Fractional CTO & Angel Investor)05:42 - From Iit Roorkee to Builder: The Pre-TCP/IP Telecom Days10:13 - The Roots of Scalability: Moving from Telecom to Paytm & Tiket10:54 - Scaling Paytm's Backend to 60,000 TPS & Navigating Black Swan Events10:14 - Managing Production SRE Teams vs. Business Velocity: The Ultimate Trade-off16:06 - The Engineering Innovation Behind Paytm’s Lightning-Fast Scan & Pay18:07 - Organizational Scale: Managing 600+ Engineers at Tiket Across 3 Countries22:05 - Matrix vs. Centralized Teams: Structuring Engineering Horizontals & Verticals24:08 - Dousing Fires vs. Fixing the Wiring: A Playbook for Tech Leadership Architecture28:25 - Founding a Healthcare Analytics Startup: The Hard Lesson of Business Viability 34:09 - The Impact of GenAI on Engineering Productivity vs. True System Design38:51- An Angel Investor's Judgment: Why AI Companies Have Zero Moat Without Distribution45:28 - The Key Trait of Successful Entrepreneurs: Long-Term Mission vs. Extreme Flexibility48:06 - Filtering Through 600+ Deals a Month: Why Being a Good Engineer Solves Any Problem50:55 - Outro & Closing Remarks: Techie Talks with Kuldeep📌 Subscribe to Techie Talks with Kuldeep for more raw engineering and leadership deep dives!------------------------------------------------------------------------CONNECT WITH US 🌐Connect with Techie Talks with Kuldeep: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 Sanjay Singh on LinkedIn:https://www.linkedin.com/in/sanjaysinghiitr/

See All (39)

About

🌟Whether you're a student exploring tech, a professional leveling up your career, or a startup founder navigating innovation, this is your space to grow. 🔥What you’ll discover here: - Learn from the best - Stay ahead in tech - Fuel your ambitions - Honest discussions about challenges, failures, mindset, growth, and learning in the tech world - Practical advice for students, job seekers, tech enthusiasts, and startup builders - Inspiring career journeys of successful tech professionals - Deep dives into cutting-edge topics Artificial Intelligence, Cloud Computing, Cybersecurity, Blockchain

Information

  • Creator
    Techie Talks with Kuldeep
  • Years Active
    2025 - 2026
  • Episodes
    39
  • Rating
    Clean
  • Copyright
    © Techie Talks with Kuldeep
  • Show Website
    Techie Talks with Kuldeep