I announced that I will speak at ContainerDays Conference 2025 in Hamburg this September, presenting a talk titled “Children's Guide to LLMs on Kubernetes” to demystify large‑language‑model inference on Kubernetes using vLLM.
containerdays.talk.title = Children's Guide to LLMs on Kubernetes
I co‑presented the session “Children’s Guide to LLMs on Kubernetes” at KubeCon + CloudNativeCon India 2025 in Hyderabad.
The goal was to break down Large Language Models for beginners and demonstrate inference at scale using Kubernetes and vLLM, proving that complex AI concepts can be approachable for everyone.
The KubeCon live demo showed a 2.3× throughput increase, reduced peak RAM from 45 GB to 13 GB, achieved 1.9× GPU scaling and 1.4× CPU scaling, and cut 95th‑percentile latency from 210 ms to 112 ms with vLLM.
These results highlight how vLLM’s KV‑batching and PagedAttention address the two biggest pain points for LLM ops: memory saturation and throughput bottlenecks, while scaling smoothly across heterogeneous CPU‑GPU clusters.
I was a panelist at the IT Conclave 2025 in Chandigarh, discussing “Agentic AI: Risks and Rewards.” The conversation highlighted how Agentic AI reshapes professional skills, acts both as a job creator and a disruptor, and drives new security‑and‑ethics measurement standards. The key takeaway was that success with AI requires balancing human judgment, creativity, and ethics with the scale and capabilities of intelligent systems.
- Event: IT Conclave 2025, PHD Chamber of Commerce & Industry, Chandigarh
The consensus was that organizations must embed human oversight and ethical frameworks alongside AI deployment to mitigate risks while unlocking benefits.
- Role: Panelist on “Agentic AI: Risks and Rewards”
- Participants: Gurdarshan Aggarwal, Shray Gupta, Anubhooti Bhatnagar, Puneet Jindal
- Demonstration VM launched with 1 NVIDIA A100 GPU attached.
- Inference latency reduced from 120 ms (CPU‑only) to 18 ms (GPU‑backed VM), a 6.7× speed‑up.
- Prometheus collected ≈ 5 k GPU‑utilization samples per minute across the cluster.
- Scheduler placed 100 % of GPU‑requested VMs onto nodes with free GPUs; zero placement failures observed.
Event dates: December 8‑10 2025 (three‑day conference) in Tokyo, Japan.
Session title: “TinyML at the Edge: Deploying and Optimizing AI Workloads on Zephyr RTOS.”
Keynote speaker: Linus Torvalds (creator of Linux).
- Calendar link: https://sched.co/29FmO
- Event page: https://lnkd.in/gnuBHPGN
- Schedule: http://bit.ly/3WkNcHe
- Registration: http://bit.ly/47nPM4o
I spoke at KCD Sri Lanka on 2025‑11‑26. The event surprised me with its energy. I thanked organizers Chamod Perera, Kemila De Silva, and Kaveesha Gimhana, highlighted opening keynote by Sanjiva Weerawarana, and listed five named keynote speakers. I also noted ten Indian community friends and eight sponsors ranging from Randoli to AWS.
The experience reinforced that community‑driven events amplify knowledge sharing and create lasting collaborations.
I was a guest panelist at the ASSOCHAM Conclave on Skill Development for MSME Sectors in Chandigarh, speaking in the session “Evolving Skill Ecosystem: Technology, Innovation & Entrepreneurship.”
The panel highlighted how MSMEs can adopt technology, AI and innovation to achieve sustainable growth and develop future‑ready skills.
OSS Japan + AI_dev + Automotive Linux Summit took place in Tokyo from Dec 8 to 10 2025 (three days).
Major sponsors included Fujitsu, NEC Corporation, ARM, Red Hat, Canonical, Amazon Web Services (AWS), SUSE, and others (eight named entities).
I decided to embrace open, safety‑certified stacks for long‑term vehicle software and to promote Zephyr RTOS as the primary OS for safe, resource‑constrained edge devices, and I will keep engaging the community to sustain momentum in open‑source GenAI, real‑time, and applied AI discussions.
I am all set for the talk The Zephyr Project ! #OSS
On 2025-12-04 I expressed excitement for the upcoming summit.
Timestamp = 2025-12-04 15:29 UTC
Message = Excited for the summit
On 2025-12-07 I announced that I will be speaking tomorrow at the Open Source Summit Japan + AI_dev: Open Source GenAI & ML Summit in Tokyo, co‑located with the Automotive Linux Summit.
Announcement timestamp = 2025-12-07 04:29 UTC
Speaker announcement = Thrilled to be speaking tomorrow at Open Source Summit Japan + AI_dev: Open Source GenAI & ML Summit here in Tokyo, co‑located with the Automotive Linux Summit.
I spoke to a student audience at GDSC CGC (Powered by Google Developers) on 2026‑01‑28, asserting that AI is not a bubble and encouraging immediate engagement.
My short answer is, No.
Thrilled to take the stage tomorrow as a panelist at the Innovation & Technology conclave organized by ASSOCHAM in Chandigarh.
assocham_panelist.announcement_timestamp = 2026-01-21 12:56 UTC
Excited to join the CXO Rabbit AI Roundtable, shaping the future of AI‑powered enterprises. I am grateful for the warm welcome from The World Organization.
Joined the CXO Rabbit AI Roundtable on 2026-02-25 20:18 UTC.
Day‑1 investment commitments = ₹10,000+ crore.
Infosys campus budget = ₹286 crore.
Infosys campus direct jobs = 2,700.
Tata Semiconductor Lab budget = ₹4,500 crore.
SGST rebate = 75% net for up to 15 years.
GCC subsidy = ₹7,500 per employee per month for first 10 units.
During the Progressive Punjab Investors’ Summit 2026 I highlighted the massive capital influx and policy incentives that position Mohali as a Tier‑1 tech hub.
At KubeCon+CloudNativeCon Europe (2026‑03‑25) I represented Welzin & CogNerd, invited fellow builders to connect, and engaged in high‑level conversations about scalable AI‑driven systems.
The event promotion mentions an awesome line up of speakers and talks.
Registrations are still open for anyone who hasn’t signed up yet.