--- title: "UJJWAL BALAJI (@ujjwal1805) | Hashnode" description: "Learning | Exploring | Currently working in Genpact as Lead Associate- EUC. Read the latest articles by UJJWAL BALAJI on Hashnode." url: "https://hashnode.com/@ujjwal1805" --- @ujjwal1805 # UJJWAL BALAJI @ujjwal1805 · Delhi, India · Joined January 2025 Learning | Exploring | Currently working in Genpact as Lead Associate- EUC [Website](https://portfolio-website-khaki-sigma.vercel.app/) ### [The AI That Refuses to Talk: Where Are We Actually Heading?](https://chatteronai.hashnode.dev/the-ai-that-refuses-to-talk-where-are-we-actually-heading?utm_source=hashnode&utm_medium=feed) TypeSafe AI's Jev skips text generation to return typed decisions in milliseconds. Here's how it works, what it costs, and where AI is heading. 15 min read · [chatteronai.hashnode.dev](https://chatteronai.hashnode.dev/the-ai-that-refuses-to-talk-where-are-we-actually-heading?utm_source=hashnode&utm_medium=feed) [1](https://hashnode.com/posts/the-ai-that-refuses-to-talk-where-are-we-actually-heading/6aaed522e5d911909597e31f) [Sep 19](https://hashnode.com/posts/the-ai-that-refuses-to-talk-where-are-we-actually-heading/6aaed522e5d911909597e31f) ### [From GPT-2 to Kimi K3: The Journey of 22,580 Models in One](https://chatteronai.hashnode.dev/from-gpt-2-to-kimi-k3-the-journey-of-22-580-models-in-one?utm_source=hashnode&utm_medium=feed) From GPT-2 to Kimi K3: Discover how AI architecture evolved from 124M to 2.8T parameters. Learn about attention mechanisms, KV cache, DeltaNet, MoE, and the innovations that made 22,580 models fit into one. Perfect for developers and AI enthusiasts. 24 min read · [chatteronai.hashnode.dev](https://chatteronai.hashnode.dev/from-gpt-2-to-kimi-k3-the-journey-of-22-580-models-in-one?utm_source=hashnode&utm_medium=feed) [Sep 6](https://hashnode.com/posts/from-gpt-2-to-kimi-k3-the-journey-of-22-580-models-in-one/6a9d5287880152a8ada16cd9) ### [Inside the Brain of an LLM: From Raw Text to Next-Token Magic](https://chatteronai.hashnode.dev/inside-the-brain-of-an-llm-from-raw-text-to-next-token-magic?utm_source=hashnode&utm_medium=feed) Dissecting how LLMs work under the hood 8 min read · [chatteronai.hashnode.dev](https://chatteronai.hashnode.dev/inside-the-brain-of-an-llm-from-raw-text-to-next-token-magic?utm_source=hashnode&utm_medium=feed) [Jun 2](https://hashnode.com/posts/inside-the-brain-of-an-llm-from-raw-text-to-next-token-magic/6a1f3648545c8a206cec4ab3) ### [Scaling PostgreSQL to 800 Million Users: OpenAI’s Surprisingly Simple Secret](https://chatteronai.hashnode.dev/scaling-postgresql-to-800-million-users-openai-s-surprisingly-simple-secret?utm_source=hashnode&utm_medium=feed) How we are scaling chat-gpt for 800 million users 7 min read · [chatteronai.hashnode.dev](https://chatteronai.hashnode.dev/scaling-postgresql-to-800-million-users-openai-s-surprisingly-simple-secret?utm_source=hashnode&utm_medium=feed) [Mar 29](https://hashnode.com/posts/scaling-postgresql-to-800-million-users-openai-s-surprisingly-simple-secret/69c8f7097816e434a01fd0c5) ### [How Amazon Used Keyboard Latency to Uncover a North Korean IT Imposter](https://ujjwal18.hashnode.dev/how-amazon-used-keyboard-latency-to-uncover-a-north-korean-it-imposter?utm_source=hashnode&utm_medium=feed) Amazon used keystroke latency data to identify a North Korean IT imposter, showcasing advanced threat detection strategies 4 min read · [ujjwal18.hashnode.dev](https://ujjwal18.hashnode.dev/how-amazon-used-keyboard-latency-to-uncover-a-north-korean-it-imposter?utm_source=hashnode&utm_medium=feed) [Dec 22, 2025](https://hashnode.com/posts/how-amazon-used-keyboard-latency-to-uncover-a-north-korean-it-imposter/69499c04f85c096bdd84950e)