Ankur Kumar
@ankurkumar.bsky.social
📝 A Techie, Blogger, Educator & Mentor
🖥️ Shares Learnings on Cloud, GenAI, Platform Engineering, Microservices, DevOps, Data Engineering
💙 Loved Husband, Proud Dad
💡 Founder Vedcraft.com - a platform for enabling Cloud, GenAI
🎯 Sports🏎️ 🎾 🏏🏒, Loves ☕🎦📘
🖥️ Shares Learnings on Cloud, GenAI, Platform Engineering, Microservices, DevOps, Data Engineering
💙 Loved Husband, Proud Dad
💡 Founder Vedcraft.com - a platform for enabling Cloud, GenAI
🎯 Sports🏎️ 🎾 🏏🏒, Loves ☕🎦📘
An illustrative maturity model (image generated with Nano Banana)
September 20, 2025 at 4:14 AM
An illustrative maturity model (image generated with Nano Banana)
✔️Berkeley Function/Tool Calling: gorilla.cs.berkeley.edu/leaderboard....
✔️LMArena: lmarena.ai/leaderboard
✔️SWE Bench: https://
✔️LMArena: lmarena.ai/leaderboard
✔️SWE Bench: https://
September 17, 2025 at 1:13 AM
✔️Berkeley Function/Tool Calling: gorilla.cs.berkeley.edu/leaderboard....
✔️LMArena: lmarena.ai/leaderboard
✔️SWE Bench: https://
✔️LMArena: lmarena.ai/leaderboard
✔️SWE Bench: https://
AI Bits - Anatomy of an AI Agent 👇
1️⃣Sensors & Perception Layer: Information about the environment, agent interacts with environment with sensors
2️⃣ Knowledge Base (aka the “Memory”) and Feedback Mechanism
1️⃣Sensors & Perception Layer: Information about the environment, agent interacts with environment with sensors
2️⃣ Knowledge Base (aka the “Memory”) and Feedback Mechanism
September 14, 2025 at 2:31 PM
AI Bits - Anatomy of an AI Agent 👇
1️⃣Sensors & Perception Layer: Information about the environment, agent interacts with environment with sensors
2️⃣ Knowledge Base (aka the “Memory”) and Feedback Mechanism
1️⃣Sensors & Perception Layer: Information about the environment, agent interacts with environment with sensors
2️⃣ Knowledge Base (aka the “Memory”) and Feedback Mechanism
3️⃣ Cost per token = capex (annualized) + energy costs + other opex/annual token output.
4️⃣Revenue = tokens X dollars per token
siliconangle.com/2025/08/30/r...
4️⃣Revenue = tokens X dollars per token
siliconangle.com/2025/08/30/r...
September 7, 2025 at 2:27 PM
3️⃣ Cost per token = capex (annualized) + energy costs + other opex/annual token output.
4️⃣Revenue = tokens X dollars per token
siliconangle.com/2025/08/30/r...
4️⃣Revenue = tokens X dollars per token
siliconangle.com/2025/08/30/r...
Gartner Technology Adoption Roadmap (‘25) for Midsize Enterprises - leverage it for planning & assessing technology strategy and roadmap for your organization 👇
✅ AI, data & analytics: GenAI, Data fabric, CDS, Metadata Management, Augmented & Prescriptive Analytics, Data Observability
✅ AI, data & analytics: GenAI, Data fabric, CDS, Metadata Management, Augmented & Prescriptive Analytics, Data Observability
June 23, 2025 at 2:11 AM
Gartner Technology Adoption Roadmap (‘25) for Midsize Enterprises - leverage it for planning & assessing technology strategy and roadmap for your organization 👇
✅ AI, data & analytics: GenAI, Data fabric, CDS, Metadata Management, Augmented & Prescriptive Analytics, Data Observability
✅ AI, data & analytics: GenAI, Data fabric, CDS, Metadata Management, Augmented & Prescriptive Analytics, Data Observability
3️⃣ Databricks Apps to securely build, deploy, and scale interactive data and AI-powered applications natively on the Databricks Data Intelligence Platform
4️⃣ MLFlow 3 for GenAI Observability
5️⃣ Delta Lake 4.0 Launched
4️⃣ MLFlow 3 for GenAI Observability
5️⃣ Delta Lake 4.0 Launched
June 11, 2025 at 11:26 PM
3️⃣ Databricks Apps to securely build, deploy, and scale interactive data and AI-powered applications natively on the Databricks Data Intelligence Platform
4️⃣ MLFlow 3 for GenAI Observability
5️⃣ Delta Lake 4.0 Launched
4️⃣ MLFlow 3 for GenAI Observability
5️⃣ Delta Lake 4.0 Launched
A Practical Guide for Architects and Developers to Build GenAI-powered Financial Applications, standardizing Contextual Data for LLMs using Model Context Protocol (MCP) 👇
medium.com/vedcraft/unl...
medium.com/vedcraft/unl...
June 3, 2025 at 3:09 AM
A Practical Guide for Architects and Developers to Build GenAI-powered Financial Applications, standardizing Contextual Data for LLMs using Model Context Protocol (MCP) 👇
medium.com/vedcraft/unl...
medium.com/vedcraft/unl...
March 29, 2025 at 3:55 PM
Observing technology trends by analysts, research companies, and thought leaders creates awareness and helps build a broader perspective. Published this article as part of my analysis - looking forward to hearing perspective from this group 👇
medium.com/vedcraft/top...
medium.com/vedcraft/top...
March 23, 2025 at 2:57 PM
Observing technology trends by analysts, research companies, and thought leaders creates awareness and helps build a broader perspective. Published this article as part of my analysis - looking forward to hearing perspective from this group 👇
medium.com/vedcraft/top...
medium.com/vedcraft/top...
DeepSeek is not just a ChatGPT competition but also a great news for the Open source community as it is the first high quality “reasoning” Open model available for fine-tuning, and self-hosting. Here is the compiled list of Open models for reference 👇
#opensource #LLMs
medium.com/vedcraft/top...
#opensource #LLMs
medium.com/vedcraft/top...
February 1, 2025 at 8:24 PM
DeepSeek is not just a ChatGPT competition but also a great news for the Open source community as it is the first high quality “reasoning” Open model available for fine-tuning, and self-hosting. Here is the compiled list of Open models for reference 👇
#opensource #LLMs
medium.com/vedcraft/top...
#opensource #LLMs
medium.com/vedcraft/top...
Agentic AI has emerged as one of the most promising GenAI building blocks in 2025, and numerous frameworks have emerged to build agentic apps. Published this article summarizing the key agentic frameworks with their architecture, features, pros/cons👇
medium.com/vedcraft/bui...
#GenAI #AgenticAI
medium.com/vedcraft/bui...
#GenAI #AgenticAI
January 25, 2025 at 4:06 PM
Agentic AI has emerged as one of the most promising GenAI building blocks in 2025, and numerous frameworks have emerged to build agentic apps. Published this article summarizing the key agentic frameworks with their architecture, features, pros/cons👇
medium.com/vedcraft/bui...
#GenAI #AgenticAI
medium.com/vedcraft/bui...
#GenAI #AgenticAI
Building AI agents for common Cloud operations accelerating organizational’s productivity and efficiency 👇
arxiv.org/html/2407.12...
arxiv.org/html/2407.12...
January 12, 2025 at 4:38 PM
Building AI agents for common Cloud operations accelerating organizational’s productivity and efficiency 👇
arxiv.org/html/2407.12...
arxiv.org/html/2407.12...
From large context window, multi-modal capabilities, accuracy, fine-tuning capabilities, prompt caching, and more - 2025 trends indicate consolidation and maturity of foundation models towards token generation efficiency, lower response time, and decreasing cost👇
blog.dataiku.com/a-dizzying-y...
blog.dataiku.com/a-dizzying-y...
January 9, 2025 at 4:48 PM
From large context window, multi-modal capabilities, accuracy, fine-tuning capabilities, prompt caching, and more - 2025 trends indicate consolidation and maturity of foundation models towards token generation efficiency, lower response time, and decreasing cost👇
blog.dataiku.com/a-dizzying-y...
blog.dataiku.com/a-dizzying-y...
NVIDIA Cosmos™ is a platform of state-of-the-art generative world foundation models (WFM), advanced tokenizers, guardrails, and an accelerated data processing and curation pipeline built to accelerate the development of physical AI systems.
www.nvidia.com/en-us/ai/cos...
www.nvidia.com/en-us/ai/cos...
January 8, 2025 at 1:11 AM
NVIDIA Cosmos™ is a platform of state-of-the-art generative world foundation models (WFM), advanced tokenizers, guardrails, and an accelerated data processing and curation pipeline built to accelerate the development of physical AI systems.
www.nvidia.com/en-us/ai/cos...
www.nvidia.com/en-us/ai/cos...
Virtual Lab AI Agent Reference Architecture by NVIDIA 👇
January 8, 2025 at 1:07 AM
Virtual Lab AI Agent Reference Architecture by NVIDIA 👇
NVIDIA partnership ecosystem
January 8, 2025 at 1:04 AM
NVIDIA partnership ecosystem
1️⃣ Blackwell architecture for accelerated computing generating thousands of tokens per second
2️⃣ AI Everywhere with enabling Windows PC as AI PC l
3️⃣ NVIDIA Agentic Framework (NeMo) and NIM Microservices with an array of AI agents for everyone
2️⃣ AI Everywhere with enabling Windows PC as AI PC l
3️⃣ NVIDIA Agentic Framework (NeMo) and NIM Microservices with an array of AI agents for everyone
January 8, 2025 at 1:04 AM
1️⃣ Blackwell architecture for accelerated computing generating thousands of tokens per second
2️⃣ AI Everywhere with enabling Windows PC as AI PC l
3️⃣ NVIDIA Agentic Framework (NeMo) and NIM Microservices with an array of AI agents for everyone
2️⃣ AI Everywhere with enabling Windows PC as AI PC l
3️⃣ NVIDIA Agentic Framework (NeMo) and NIM Microservices with an array of AI agents for everyone
That’s very well articulated article on Small Language Models along with:
1️⃣ Compression methods: Pruning, Quantization, Low-rank factorization, Knowledge distillation
2️⃣ Popular SLMs:
● DistilBERT
● Gemma
● GPT-4o mini
● Granite
● Llama
● Ministral
● Phi
#LLM #SLM #GenAI
www.ibm.com/think/topics...
1️⃣ Compression methods: Pruning, Quantization, Low-rank factorization, Knowledge distillation
2️⃣ Popular SLMs:
● DistilBERT
● Gemma
● GPT-4o mini
● Granite
● Llama
● Ministral
● Phi
#LLM #SLM #GenAI
www.ibm.com/think/topics...
December 21, 2024 at 4:46 PM
That’s very well articulated article on Small Language Models along with:
1️⃣ Compression methods: Pruning, Quantization, Low-rank factorization, Knowledge distillation
2️⃣ Popular SLMs:
● DistilBERT
● Gemma
● GPT-4o mini
● Granite
● Llama
● Ministral
● Phi
#LLM #SLM #GenAI
www.ibm.com/think/topics...
1️⃣ Compression methods: Pruning, Quantization, Low-rank factorization, Knowledge distillation
2️⃣ Popular SLMs:
● DistilBERT
● Gemma
● GPT-4o mini
● Granite
● Llama
● Ministral
● Phi
#LLM #SLM #GenAI
www.ibm.com/think/topics...
That’s interesting that Langfuse team didn’t think about caching prompts and using OLAP DB in early stages 🤔
Lessons learned: launch your product quickly and as it scales, evaluate and scale your architecture 👍🏻
langfuse.com/blog/2024-12...
Lessons learned: launch your product quickly and as it scales, evaluate and scale your architecture 👍🏻
langfuse.com/blog/2024-12...
December 19, 2024 at 2:06 AM
That’s interesting that Langfuse team didn’t think about caching prompts and using OLAP DB in early stages 🤔
Lessons learned: launch your product quickly and as it scales, evaluate and scale your architecture 👍🏻
langfuse.com/blog/2024-12...
Lessons learned: launch your product quickly and as it scales, evaluate and scale your architecture 👍🏻
langfuse.com/blog/2024-12...
The definition of golden path differs for different enterprises and so start your journey with an internal reference architecture for platform engineering to improve the development productivity incrementally.
platformengineering.org/blog/how-to-...
#platformengineering
platformengineering.org/blog/how-to-...
#platformengineering
December 10, 2024 at 3:17 AM
The definition of golden path differs for different enterprises and so start your journey with an internal reference architecture for platform engineering to improve the development productivity incrementally.
platformengineering.org/blog/how-to-...
#platformengineering
platformengineering.org/blog/how-to-...
#platformengineering
Community initiatives like OPEA is great to build standards and solutions together 👍🏻
OPEA (Open Platform for Enterprise AI) is a framework that enables the creation and evaluation of open, multi-provider, robust, and composable generative AI (GenAI) solutions.
opea-project.github.io
OPEA (Open Platform for Enterprise AI) is a framework that enables the creation and evaluation of open, multi-provider, robust, and composable generative AI (GenAI) solutions.
opea-project.github.io
December 8, 2024 at 6:45 PM
Community initiatives like OPEA is great to build standards and solutions together 👍🏻
OPEA (Open Platform for Enterprise AI) is a framework that enables the creation and evaluation of open, multi-provider, robust, and composable generative AI (GenAI) solutions.
opea-project.github.io
OPEA (Open Platform for Enterprise AI) is a framework that enables the creation and evaluation of open, multi-provider, robust, and composable generative AI (GenAI) solutions.
opea-project.github.io
Going to check this framework out for AI apps development with LLMs for agentic workflows 👇
llmware.ai
llmware.ai
December 8, 2024 at 4:11 AM
Going to check this framework out for AI apps development with LLMs for agentic workflows 👇
llmware.ai
llmware.ai
AWS re:Invent '24 Takeaways from Keynote (Matt Garman) - Nova, S3, Q and Bedrock announcements are not to be missed 🧵
#aws #awsreinvent
#aws #awsreinvent
December 4, 2024 at 4:54 PM
AWS re:Invent '24 Takeaways from Keynote (Matt Garman) - Nova, S3, Q and Bedrock announcements are not to be missed 🧵
#aws #awsreinvent
#aws #awsreinvent
Interesting apps like msty.app to use LLMs locally will rise for local development with RAG support. But need more open source solutions with Apache/MIT license (Mstt license is restricted for personal usage).
December 1, 2024 at 6:03 PM
Interesting apps like msty.app to use LLMs locally will rise for local development with RAG support. But need more open source solutions with Apache/MIT license (Mstt license is restricted for personal usage).