Writing & notes
On AI, product design, growth, and the things I’m still figuring out. Arguments, experiments, and notes from the work.
20 pieces of writingI recently finished Rolf Dobelli’s The Art of Thinking Clearly, 99 short chapters on the ways our minds fool us. It was remarkably… Continue reading on Medium »
technology · ai · llm · startup
Why traditional credit scores fail fast-growing businesses, and how we modeled daily cash volatility with Z-scores to deploy millions with zero initial defaults.
Fintech · Risk Modeling · Underwriting · Python
From fragile Ruby prompts to a typed LangGraph state machine with fail-closed Composio authorization for loan document classification and policy validation.
LangGraph · Bedrock AgentCore · Agentic Systems · Mortgage Tech
How we combined dense vector embeddings (Pinecone) with sparse lexical search (Elasticsearch) to reduce manual bookkeeping categorization by ~80%.
Pinecone · Elasticsearch · Hybrid RAG · Django
Where protocol state belongs, why request boundaries matter, and what to check before migrating an MCP server.
MCP · AI Protocols · Systems Design
Follow a message through identity, permissions, an ACP bridge, and agent tools. What Buzz teaches us about collaborative human-agent workspaces.
Nostr · Agent Protocols · Product Design
Writing · AI · Philosophy · Ideas
Reflection · Writing · Philosophy
Reflection · Writing · Philosophy
Ideas · Projects · Todo
Philosophy · Writing
Writing · Learnings · Reflection
Codebase · Testing · Guides and Processes · Infrastructure
Empirical study on Socratic guardrails and learning retention for CS students. Published at the 2024 IEEE Frontiers in Education Conference (FIE).
IEEE FIE 2024 · Empirical Study · CS Pedagogy
In this blog post we will deploy Microsoft Phi-3 instruct model using vLLM, an open source library designed for efficient LLM inference… Continue reading on Generative AI »
deployment · gpu · open-source · startup
Master's thesis research at Virginia Tech (4.0/4.0 GPA, 130+ citations). Exploring adaptive CS learning environments with Socratic guardrails and RAG.
Master's Thesis · Socratic AI · CS Education · Virginia Tech
Meta just released their latest Llama3 model. With every breakthrough, the ecosystem of Open Source models is getting stronger and… Continue reading on Generative AI »
artificial-intelligence · data-science · machine-learning · large-language-models
I have always believed automation to be the way in the actualization of Large Language Models. Reducing human error in organization and… Continue reading on Medium »
artificial-intelligence · engineering · machine-learning · technology
In the realm of Software Development and by extension in Machine Learning, version control systems are indispensable tools that enable… Continue reading on DevOps.dev »
devops · software-development · mlops · project-management
Dawn of the Planet of Language Models Continue reading on Medium »
large-language-models · interview · vector-database · artificial-intelligence
I always find it difficult to use Interfaces to read something. I have adopted the ‘reading between the lines’ method but it does not help… Continue reading on Medium »
machine-learning · software-development · ai · engineering
Lets build something quick and fun! Continue reading on Medium »
ai · llm · family · movies
Stable Code 3B is a 2.7 Billion parameter Language Model allowing accurate and responsive code completion at a level on par with models… Continue reading on Medium »
artificial-intelligence · code · open-source · large-language-models
Published with anticipation for the transformative era that Llama 3 will usher in. Continue reading on Stackademic »
large-language-models · agi · technology · startup
Full-stack productivity application uniting task execution, focus blocks, and telemetry.
Productivity · Full Stack · TypeScript
Comprehensive research survey on explainability and trust in AI-driven pedagogical systems. Published at IEEE SoutheastCon 2023 with 106+ citations.
IEEE Publication · Explainable AI · Citations: 106+
Distributed web search and summarization engine combining inverted index postings with neural sequence summarization.
Information Retrieval · Transformers · NLP