Rithindatta Gundu (@RithinDatta) header
Rithindatta Gundu (@RithinDatta) profile picture

AI Engineer @USTglobal | Tech Enthusiast | Always coding, Always learning. Let's connect!

Redmond, WA
Joined June 2015
Photo by Rithindatta Gundu (@RithinDatta)Photo by Rithindatta Gundu (@RithinDatta)Photo by Rithindatta Gundu (@RithinDatta)Photo by Rithindatta Gundu (@RithinDatta)Photo by Rithindatta Gundu (@RithinDatta)Photo by Rithindatta Gundu (@RithinDatta)Photo by Rithindatta Gundu (@RithinDatta)Photo by Rithindatta Gundu (@RithinDatta)Photo by Rithindatta Gundu (@RithinDatta)Photo by Rithindatta Gundu (@RithinDatta)
Microsoft IQ is the Context Gap as a product: governed context layer across Work, Fabric, Foundry and Web IQ. Same models, different context. #AIAgents #ContextGraph
8
GraphRAG Zero is live. Graph-based retrieval is finally ready for production agents.
GraphRAG Zero solves a fundamental challenge that has limited AI agents' ability to reach their full potential. It’s a technology for holistic discovery that reflects years of Microsoft research, transforming graph-based retrieval from an experimental technique into a deployable system. Optimized for real-world applications and agents, it enables organizations to explore connections across vast amounts of their own data, quickly and at low cost. It’s available now in Foundry IQ and @github’s Repo Mind Light, supporting applications from scientific discovery to repository understanding. msft.it/6017xE5xL
18
Rithindatta Gundu retweeted
we need a better productivity metric than PRs merged anyone had any success measuring business or user value created by those PRs?
114
4
4
240
29,640
Rithindatta Gundu retweeted
Bring Jensen back
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me. I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA. I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation. Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO. Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours. Jensen
163
423
19
7,773
128,477
Rithindatta Gundu retweeted
Introducing MAI Code 2 and Code 2 Flash. Code 2 brings frontier price-performance for everyday enterprise coding tasks. Code 2 Flash brings affordable and high-quality coding for routine work at scale. Try both today in GitHub Copilot.
50
107
45
1,053
71,498
Rithindatta Gundu retweeted
#SenseNova-Loop-9B is officially here — now #OpenSource! Just 9B parameters, competitive with Qwen3.8-27B across multiple benchmarks. 🔹 Big gains over Preview: NL2Repo 19.91 → 41.36; AutomationBench 14.33 → 36.67; GAIA 37.86 → 69.90 🔹 Loop + mHC: 32 unique layers, 56 layer executions 🔹 Day-0 4-bit AWQ weights (~4.9GB), with VRAM needs as low as ~9 GiB at 64K context 🔹 Up to 512K context, with text and image inputs See what a 9B model can build: animated posters, interactive websites and 3D scenes. 💻 GitHub: github.com/OpenSenseNova/Sen… 🤗 HF: huggingface.co/collections/s… 👾 Discord: discord.com/invite/BuTXPHmQu…
48
156
70
1,259
117,188
Rithindatta Gundu retweeted
$AMD chiplets is the future! 📶 Long term, I win! Better yield Lower cost More energy efficient Flexible integration As scaling a single physical die becomes economically and physically harder, chiplets and advanced packaging (like 2.5D/3D stacking and standards like UCIe) let designers scale performance horizontally and vertically. Not Financial Advice! DYOR!
Why $AMD is going to >2x from here 2027-2028 🧵 Real reasons why Institutions are FOMOing into @AMD regardless of SI sentiment Not Financial Advice! DYOR! TLDR: Hyperscalers, SI Labs and Enterprises is allocating a meaningful % of total CapEx to CPUs in 2027 toward the next 10 years, because of Consumer and Enterprise Agentic. In short, CPUs will generate far higher revenue growth and profitability for the major players! Higher revenue/profitability growth = more CapEx Not less FY2027 Estimate/Projection: Revenue $100B+ NonGAAP EPS: $22-$25 ~Agentic SI is a real inflection that is pulling more CPU cores and changing historical CPU:GPU ratios toward higher CPU shares in Agentic workloads. ~This supports stronger growth in agent related revenue for hyperscalers and SI labs by making the overall stack more productive and cost efficient. ~It does not incentivize lower CapEx, but higher due to higher revenue growth and profitability short-medium-long term. 1. Institutions FOMOing will drive $AMD price action much higher due to tiny float vs other semis. And Agentic structurally changes how CapEx is going to be allocated from here! Agentic SI is structurally rewriting CapEx allocation away from pure GPU dominance toward high density CPUs, early phase is estimate to transition toward 4-8 CPU 1 GPU ratio from 1 : 1 recently change. AMD is positioned to capture a disproportionate/highest share of that shift through EPYC Venice and the Helion rack blueprint, even amid broader semiconductor skepticism over ROI, power, and financing. Traditional generative AI (chatbots, single turn inference) favored GPU heavy racks at roughly 1:4 to 1:8 CPU:GPU ratios. Agentic systems are different. A single user request triggers multi step planning, tool calls, code execution, retrieval (RAG), database/API interactions, policy checks, memory management, and orchestration across sub agents. These are sequential, branch heavy, latency sensitive workloads that run primarily on CPUs. Studies and vendor data show CPU side orchestration can account for 50-90% of total latency in agentic workflows. Adding more GPUs does not fix the bottleneck if the control plane is starved. This drives a ratio shift toward 4-8 CPU : 1 GPU (or higher on the CPU side). $ARM estimates CPU core demand per gigawatt of data center capacity rising from ~30 million to ~120 million (4×). AMD’s Dr. Lisa Su has described the move from classic 1:4/1:8 host-node configurations toward 1:1, with the possibility of more CPUs than GPUs at high agent density and upgraded CPU TAM to $220B+ by 2030 after 3-4 revisions. @intel CEO has made similar comments about near parity or higher CPU counts. Morgan Stanley estimates agentic AI adds $32.5–60 billion of incremental CPU TAM by 2030 on top of a server CPU market already expected to exceed $100 billion (some frameworks push the total server CPU TAM toward $80–110B+). Memory demand rises in parallel. Hyperscalers ($MSFT $META $AMZN $GOOGL ), sovereign AI operators, SI labs (@OpenAI @AnthropicAI ), and enterprises therefore allocate a meaningfully higher percentage of CapEx to CPUs. Total Big 4 hyperscaler CapEx is guided in the $720–740B range for 2026 and higher for 2027, with servers/chips as the largest line item and further growth expected into 2027. GPU racks remain essential for model inference and training, but the surrounding agent execution layer scales with tokens and concurrent agents exactly the token hungry, high volume regime of production agentic systems. CPU dense capacity improves GPU utilization (by keeping accelerators fed) and enables cheaper orchestration of mixed frontier/open source model calls. AMD is the biggest merchant beneficiary on the CPU side. Its 6th gen EPYC “Venice” (Zen 6, up to 256 cores / 512 threads on SP7, shipping) delivers strong performance per core, per watt, and per dollar. Data Center revenue already hit a record ~$8.5B in Q3 2026 (+97% YoY, ~64% of total company revenue). Management expects server (CPU) revenue to roughly double in 2027, Data Center to exceed 80% of total revenue, and company revenue to grow "well over 100%" for data center segment. Venice ramps plus the Helion architecture specifically address the agentic density problem from November 2026. 2. AMD Helion/CPU Rack vs GPU rack comp Hyperscale / CPU only: Up to 128 × EPYC 9996 class processors → 32,768 Zen 6 cores (>65,536 threads). Configurable dense (256 core/400 W), balanced (256 core/600 W), or high frequency (96-core/~5 GHz) blades matched to personal, enterprise, or technical agents. Connects to separate inference infrastructure. Up to 102 Tbps networking. Liquid or air cooling. Integrated token router directs calls across frontier and local open source models. Dense liquid cooled Venice CPU only Helion (128 × 256-core class): $1.8–$3 million (est) Mixed Venice Helion Rack: $1.2-$2M Air Cooled lower density =$700k-$1.5M vs GPU racks are in the range of $6-$10M/Rack atm AMD CPU rack is 30%-89% than most GPU racks. CPU silicon is far less expensive per unit of relevant compute than high end GPUs. A dense 128 CPU rack has CPU content on the order of ~$1.9M at list (EPYC 9996 ~$14,904 1kU), before memory, networking, chassis, and cooling realistically a few million dollars fully configured versus $6-$10M GPU racks. Higher agents per watt and agents per dollar for the orchestration layer. Net effect on CapEx mix: For every dollar of GPU inference capacity, operators now need substantially more CPU capacity to run the agents that generate and orchestrate the token demand. That raises the CPU percentage of total server CapEx relative to the pure generative SI era. AMD captures both sides (EPYC in Helion and as Helios hosts + Instinct), but the volume and margin leverage from high core count Venice CPUs is especially powerful because the addressable CPU layer is expanding faster than previously modeled. EPYC also has much higher margin than GPU. Conclusion: Dr. Lisa Su has consistently argued that agentic workloads would become a major new driver of cloud CPU demand, and the data is confirming that view. She has described SI agents as the “next killer application” for cloud CPUs because multi step agent systems spend significant time on orchestration, tool calling, code execution, state management, sandboxing, and coordination between model inferences. These tasks are sequential or lightly parallel and map far better to high core count CPUs than to GPUs. This is exactly why hyperscalers and SI labs are raising CPU to GPU ratios and placing large CPU orders (Meta’s multi year deal for millions of AMD EPYC to scale @Muse is a clear example). Yole Group’s recent tracking shows data center CPU revenue growing faster in percentage terms than GPUs in recent quarters, driven by this agentic demand, while AMD’s own results have shown robust EPYC server CPU growth even as Instinct GPU revenue ramps. Su’s point was not that CPUs would displace GPUs, but that the agentic wave would create a large, sustained incremental CPU opportunity that improves overall system utilization and economics. That prediction has held up. Not Financial Advice! DYOR!
4
4
64
3,622
Rithindatta Gundu retweeted
Built for developers. Ready for local AI. At Advancing AI 2026, Jack Huynh shared why AMD Ryzen AI Halo was designed to help developers build, test, and iterate on AI models locally.

3
5
43
188,058
Rithindatta Gundu retweeted
Today, we are launching Sarvam Token Factory for everyone. We host our own models, and the full range of best-in-class open-weight models, in India. We match leading inference providers on performance and reliability. For enterprises, this means more models to choose from, and an AI stack they can own without lock-ins.
255
368
321
3,134
1,080,735
Rithindatta Gundu retweeted
AI ENGINEERING FROM SCRATCH
1
95
1
869
24,226
Rithindatta Gundu retweeted
Over the past week, I’ve been overwhelmed by the love and support of friends, industry colleagues, and people I've never met. Thank you to everyone who has shown kindness to my family and me. It means the world to me. I came to the United States 53 years ago, and becoming an American citizen is one of the greatest gifts and proudest moments of my life. NVIDIA has been my life’s work. Building it into the company it is today alongside my colleagues has been a privilege and an honor. Nowhere but in America could a company like NVIDIA be possible. I love our country, and I love NVIDIA. I did not live up to the standards expected of everyone at NVIDIA, and I have made serious mistakes in my personal life. I created problems for the Board that I deeply regret. I own my mistakes completely, and I am committed to cooperating with the Board’s independent investigation. Since my departure, I’ve seen an outpouring of calls from NVIDIA’s employees, investors, business partners, and customers for me to return as CEO. Many continue to reach out to me directly as the company navigates complex technology, market, competitive, and geopolitical dynamics. NVIDIA is a once-in-a-generation company. If the Board of Directors decides that my return is in the best interest of our shareholders, our employees, the company, and America, I would be honored to answer these calls and return as CEO. Thank you again to everyone who has shared their good wishes. I am profoundly humbled by your support. To my family: I love you, and I will spend the rest of my life being worthy of yours. Jensen
59,382
120,381
24,273
1,153,717
55,301,882
Rithindatta Gundu retweeted
🚀 Big news! Neo4j Virtual Graph is now generally available in Neo4j Aura on GCP! You can now build zero-ETL knowledge graphs directly on top of your existing warehouse tables in Snowflake, Databricks, and Google BigQuery without copying or migrating data. This release includes new features, such as the ability to ground your Aura Agents with context from your data, run algorithms in graph analytics sessions using Cypher, build dashboards with AI, and more. With Virtual Graph on GPC, keep your governance and benefit from a graph. More? bit.ly/4C7BdAn #AI #Virtualgraph #GPC
3
6
38
2,863
Rithindatta Gundu retweeted
10 things to know before joining the AMD AI Developer Program 👇 1. It’s free. Anyone can join, whether you are a student starting out or a seasoned ML engineer. 2. You get $100 in cloud credits on day one. Enough to prototype, benchmark, and get moving on AMD tech. 3. You don’t need any hardware to start. Build immediately with cloud access. 4. There’s a private Discord with a community of engineers who will provide technical discussion and help. 5. You get early invites to dev events and workshops. Before they’re public. Before they fill up. 6. Monthly Ask the Expert hours with open-source partners. Ask questions. Go deep. Learn the tricks of the trade. 7. Your projects can get featured on official AMD Developer channels and at events. 8. You’re automatically entered into monthly hardware raffles. No forms. No extra steps. 9. You get a free month of @DeepLearningAI Pro. High quality courses for you to expand your skills. 10. The more you participate, the more you unlock. More access, more visibility, more opportunities. If you’re building, training, or optimizing AI models, we’d love to have you join us!
24
82
5
710
38,709
Rithindatta Gundu retweeted
One click. Hands-on with AMD GPUs. Five AMD Developer Hub tutorials now launch directly in AMD Developer Cloud on AMD Instinct MI300 GPUs, with preconfigured environments ready to run. These can now be launched in a single click: - AI agent with MCPs using vLLM and PydanticAI (Inference) - Running ComfyUI generative workflows from Python on AMD Instinct GPUs (Inference) - OCR with vision-language models with vLLM (Inference) - Fine-tune Llama-3-8B with Llama-Factory (Fine-tuning) - Kernel development and optimization with Triton (GPU Development & Optimization) Choose a tutorial and click “Open in AMD Developer Cloud” to get started: rocm.docs.amd.com/projects/a… AMD Developer Cloud sign-in required.
7
8
1
60
6,066
Rithindatta Gundu retweeted
Is anyone actually using Jev? 1mo ago the hype was amazing; if you search on YouTube you find videos from then, but none now. I tried it for many classification jobs and it was hard to get it just to match Sonnet.
558
21
33
2,149
358,160
Rithindatta Gundu retweeted
AI coding agents can waste time and tokens relearning the same lessons. TokenEconomizer builds on past sessions to cut costs while maintaining task success. It’s a path toward coding agents that learn not only how to solve tasks, but also how to spend computation more effectively. The open-source plugin is available now on @github. msft.it/6017xDtyc
41
88
18
886
54,333
Rithindatta Gundu retweeted
We're teaming up to accelerate open-source Al. OpenRouter is now integrated into Hugging Face directly. Same token, same endpoint, plus live model stats on the Hub and an inference widget we redesigned together. The best of open, in one place. huggingface.co/blog/openrout…
201
334
117
2,680
121,238
Rithindatta Gundu retweeted
Say no to crack
7,581
9,293
1,194
161,268
5,545,562
Rithindatta Gundu retweeted
Congratulations to an incredible team at SpaceX!!!
7,061
17,709
2,156
169,714
7,728,583