<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Ai on Ash Tewari</title><link>https://www.tewari.info/tag/ai/</link><description>Recent content in Ai on Ash Tewari</description><generator>Hugo</generator><language>en-US</language><lastBuildDate>Tue, 20 Oct 2026 01:52:51 +0000</lastBuildDate><atom:link href="https://www.tewari.info/tag/ai/feed.xml" rel="self" type="application/rss+xml"/><item><title>All Things Open 2026 - Session Recommendations</title><link>https://www.tewari.info/2026/10/19/all-things-open-2026-session-recommendations/</link><pubDate>Tue, 20 Oct 2026 01:52:51 +0000</pubDate><guid isPermaLink="false">https://www.tewari.info/?p=1591</guid><description>&lt;p&gt;&lt;a href="https://2026.allthingsopen.org/"&gt;All Things Open 2026&lt;/a&gt; is on in Raleigh! With more than 100 sessions and speakers, choosing what to attend is not easy.&lt;/p&gt;
&lt;p&gt;I went through the conference program and shortlisted the sessions that caught my attention. My interests lean toward AI agents, software architecture, developer productivity, security, and data engineering, practical engineering challenges, along with a few interesting topics on model training, real-time AI, and emerging user interfaces.&lt;/p&gt;
&lt;p&gt;AI coding assistants have evolved considerably. We&amp;rsquo;re now moving toward autonomous agents that can modify repositories, execute commands, provision infrastructure, and even deploy applications. But giving an agent the ability to write code is relatively easy. Making it reliable, secure, efficient, and useful in a real engineering environment is a much harder problem.&lt;/p&gt;</description></item><item><title>Agentic finetuning using a custom harness</title><link>https://www.tewari.info/2026/09/24/agentic-finetuning-using-a-custom-harness/</link><pubDate>Fri, 25 Sep 2026 02:56:11 +0000</pubDate><guid isPermaLink="false">https://www.tewari.info/?p=1557</guid><description>&lt;p&gt;The goal was to figure out whether a small model running on my own hardware could do the job of intent classification and structured parameter extraction  well enough for one of my use-cases - read a request like &amp;ldquo;add 8 spare tent stakes to the camping box&amp;rdquo; and turn it into something the application can act on.&lt;/p&gt;
&lt;p&gt;I created an agentic fine-tuning and evaluation loop with a custom evaluation harness: an AI agent proposes one change at a time, a custom evaluation harness measures it against a frozen test set, and every run gets appended to a log with its settings, result and next decision. The agent does the all the work of dataset generation, training, evaluation and diagnosis.&lt;/p&gt;</description></item><item><title>Giving OpenClaw Secure Access to Cloud Services Without Sharing Your Password</title><link>https://www.tewari.info/2026/04/06/device-code-auth-flow-with-openclaw/</link><pubDate>Mon, 06 Apr 2026 04:48:17 +0000</pubDate><guid isPermaLink="false">https://www.tewari.info/?p=1498</guid><description>&lt;p&gt;Device Code Flow has been built into OAuth2 for years, originally designed for TVs and game consoles. It works just as well for a Docker container. It requires no credentials stored on the agent machine. It gives you narrow, revocable access that the agent cannot exceed.&lt;/p&gt;
&lt;p&gt;I have OpenClaw deployed in a sandboxed docker container on a dedicated host. I communicate with it via a Telegram bot, in a locked-down chat.&lt;/p&gt;</description></item><item><title>Understanding RAG @ All Things Open AI 2025</title><link>https://www.tewari.info/2025/03/30/understanding-rag-all-things-open-ai-2025/</link><pubDate>Mon, 31 Mar 2025 02:29:39 +0000</pubDate><guid isPermaLink="false">https://www.tewari.info/?p=1468</guid><description>&lt;p&gt;&lt;strong&gt;What I learned while creating Bookshelf (open source RAG Application)&lt;/strong&gt;&lt;/p&gt;
&lt;p&gt;I had the opportunity to speak at the &lt;a href="https://allthingsopen.ai/"&gt;AllThingsOpen.ai 2025 Conference&lt;/a&gt; recently. It was a wonderful experience to learn from and share with so many amazing fellow presenters and attendees. All Things Open AI 2025 session recordings are available now.&lt;/p&gt;
&lt;p&gt;I spoke about several key aspects of Retrieval-Augmented Generation (RAG) and vector databases. The main topics I addressed included The Core Concepts of RAG, Embedding Models vs. Inference Models, Handling Multiple Embeddings and various Optimizations Beyond Naive RAG such as chunking strategies, enrichment with metadata, auto-merging retriever, and reranking models. Finally, I touched on Local Execution and GPU Utilization.&lt;/p&gt;</description></item><item><title>SPCHR: Speech To Text App for Windows</title><link>https://www.tewari.info/2025/02/09/spchr-speech-to-text-app-for-windows/</link><pubDate>Mon, 10 Feb 2025 04:32:25 +0000</pubDate><guid isPermaLink="false">https://www.tewari.info/?p=1456</guid><description>&lt;p&gt;I&amp;rsquo;m excited to announce the release of SPCHR. Pronounced as &amp;ldquo;speaker&amp;rdquo;. I know, I know. My apologies, it was too late at night and that&amp;rsquo;s the best I could come up with ;)&lt;/p&gt;
&lt;ul&gt;
&lt;li&gt;It is a Windows desktop application for speech-to-text transcription.&lt;/li&gt;
&lt;li&gt;You can use it to &amp;ldquo;add&amp;rdquo; voice input to applications that don’t natively support it. It works anywhere you can type or paste text.&lt;/li&gt;
&lt;li&gt;You can use it entirely locally on your PC if you like, so it is private and secure.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="key-features"&gt;Key Features&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;&lt;strong&gt;Flexible&lt;/strong&gt;: SPCHR can use either Azure Speech Services or a local OpenAI Whisper model, giving you the flexibility to choose between cloud-based and local transcription.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Zero Configuration&lt;/strong&gt;: The application works immediately with local Whisper processing - no account setup required!&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Global Hotkey&lt;/strong&gt;: Start and stop recording from &lt;em&gt;any&lt;/em&gt; application with Ctrl+Alt+L.&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Seamless Integration&lt;/strong&gt;: Transcribed text is automatically pasted into your active window.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="getting-started"&gt;Getting Started&lt;/h2&gt;
&lt;ul&gt;
&lt;li&gt;Clone the repository&lt;/li&gt;
&lt;li&gt;Build the solution using Visual Studio and run it&lt;/li&gt;
&lt;li&gt;&lt;strong&gt;Press Ctrl+Alt+L and start speaking&lt;/strong&gt;&lt;/li&gt;
&lt;li&gt;Watch as your words appear in your active window&lt;/li&gt;
&lt;li&gt;For those wanting cloud-based transcription, simply add your Azure Speech Services credentials to the configuration file.&lt;/li&gt;
&lt;/ul&gt;
&lt;h2 id="open-source"&gt;Open Source&lt;/h2&gt;
&lt;p&gt;SPCHR is released under the MIT License, and I welcome contributions from the community. Whether you&amp;rsquo;re interested in adding features, fixing bugs, or improving documentation, your help is appreciated.&lt;/p&gt;</description></item><item><title>WhichBox - AI Assistant App</title><link>https://www.tewari.info/2024/08/04/whichbox-ai-assistant-app/</link><pubDate>Sun, 04 Aug 2024 17:37:56 +0000</pubDate><guid isPermaLink="false">https://www.tewari.info/?p=1427</guid><description>&lt;p&gt;Do you remember which one of those boxes in the garage has your old phone? Or that toy your child wants to play with again? Or the tripod you need for your weekend trip? Oh what about that massager that you could really use right now! You would probably have to dig through those boxes to find it. Same problem when moving. There are always some boxes that you don&amp;rsquo;t want to open right away, but you wish you could open just the right one when you need something.&lt;/p&gt;</description></item></channel></rss>