ML Finance seminar today: sites.google.com/view/mlfinan...égoire-macqueron-jan-14
14.01.2026 09:36
👍 0
🔁 0
💬 0
📌 0
ML Finance seminar today: sites.google.com/view/mlfinan...égoire-macqueron-jan-14
A new book: "Deep Learning in Quantitative Trading"
www.cambridge.org/in/universi...
Code available here: github.com/zcakhaa/Dee...
#CBOE #VIX Index Decomposition Tool
The August 2024 spike and this year's "Liberation Day" in April appeared similar. However, sentiment remained notably more bullish during the April spike.
White paper:
www.cboe.com/insights/pos...
A screenshot essay titled 'The power of MCP on the desktop' discussing the author's use of the Claude desktop app on a Mac for project management with Linear App. The text highlights the integration of these tools for efficient workflows, including the creation of a specific project in Claude Desktop with instructions for managing the Linear workspace. The essay mentions functions like asking Claude to flag issues that lack descriptions, estimate timelines, amend existing issues, and add new ones. The author emphasizes that, despite these tools, actual work still requires effort and cannot be shortcut.
How I use the Claude desktop app with Linear App.
Pretty good for in-flight WiFi: 44M down. Finnair HEL-LHR.
Screenshot essay titled '73GB LLMs on iPhone' discussing how to run private AI models on an iPhone using a 3-tool setup. The text highlights the need for accessing custom AI models away from a computer, emphasizing the benefits of using Ollama on a Mac for local model running, Tailscale for a private network tunnel, and the LLM Bridge app for mobile connectivity. It includes an icon of a computer next to Ollama, a link icon next to Tailscale, and a mobile phone icon next to the LLM Bridge app. The essay explains how enabling 'Expose Ollama to Network' transforms an iPhone into a gateway for AI research, with a focus on privacy, mobility, and customization. A link to the LLM Bridge app on the Apple website is also provided.
The way I use Ollama LLMs on my iPhone
Hi
Apologies for the late reply. I was offline for a few days.
I'd say it's mostly two things
- it gives me a quick start into a large number of papers.
- perhaps most importantly, it's the connections between papers. I rely heavily on Obsidian, with its links between documents.
Trading agent. Paper and code.
arxiv.org/abs/2412.20138
code here:
github.com/TauricResea...
I was able to get this up and running in < 10 minutes!
A screenshot essay titled 'Why Cursor Web Agent Redefines Mobile GitHub Workflows' discusses the author's experience using the Cursor Web Agent app. The text describes tasks related to coding that can be done easily without assistance. The author shares a specific morning task where they used the Cursor app on their phone to rename a function, update documentation, and create a pull request, which they later reviewed on their desktop. The essay highlights the asynchronous design of Cursor Web Agent, which allows for background operation without constant interaction, and its simplicity and effectiveness in handling tasks with minimal instructions. The author is in the early stages of exploring the tool.
#Cursor Web Agent for #coding.
Stoikov market making using the Hummingbot trading engine @_hummingbot.
papers.ssrn.com/sol3/papers...
Unfortunately, the paper does not include code. Hummingbot of course already have an earlier implementation of the Avellaneda & Stoikov paper: hummingbot.org/strategies/...
A screenshot essay titled 'Your Zotero Library Just Got Superpowers' discusses the transformation of Zotero into an AI-powered research tool. The author describes eliminating the burden of reading numerous PDFs by analyzing an entire library in minutes with AI models like Claude, GPT-4, and local Llama models. They explain how the system integrates with Zotero's database, utilizing existing tags as search filters and maintaining citekeys while linking with Obsidian notes. The author shares their experience using the tool to analyze podcast transcripts, focusing on extracting actionable insights such as trading strategies. The piece emphasizes the potential of AI to enhance knowledge work, positioning the Zotero library as a personal research engine rather than mere storage. The author hints at a more detailed Medium post to come.
How I use #LLMs to tap into my #Zotero database of papers, academic or otherwise.
#Tools I'm considering: #Ulysses for writing. Like many advanced tools, it has a learning curve. Its Markdown-based, distraction-free writing environment is very appealing.
#obsidian is my go-to #PKM tool due to its potent simplicity. Pair it with desktop #claude and #MCP integration, and you're off to the races. I'll write a Medium article on this, including how to use Langchain for even more powerful workflows.
Books I'm reading: Hands-On AI Trading. Using @QuantConnect for AI-driven analysis and development. Searching for useful insights.
Just like everybody else, I now have a copy of this new gem: the latest by @__paleologo
Violin Concerto No. 2 by Florence Price, performed by the virtuoso Randall Goosby at LAPhil. Just wow - What a gem!
A comprehensive overview of agents (264 pages!)
arxiv.org/abs/2504.01990
The timeline is mind-boggling: a lot will happen in the next two years.
ai-2027.com
A thought-provoking interview on #AI:
Sam Altman at TED:
www.youtube.com/watch?v=5MW...
Audain Art Museum in Whistler, BC. A wonderful collection of First Nations art.
Indeed it was! I had forgotten to get tickets but your comment helped.
A great recent update to @QuantRocket_LLC : Apple Silicon, including @IBKR API now fully supported. Running in Docker locally is very powerful, when you have lots of RAM.
Tools of the trade for running LLMs locally: the combination of @ollama and @OpenWebUI is very powerful. Run most recent models, including #Deepseek-r1 @deepseek_ai .
Stream Deck! I decided to join this rarefied group of desktop magicians who use this device. Went straight for the XL...
Carol Alexander on Order Flow Impact in crypto #QuantMinds
Graph neural networks for time series prediction
Richard Turner
Mesirow
#QuantMinds
Stefano Pasquali
BlackRock
Multi-agent architectures for modelling competitive trading strategies in electronic markets
#QuantMinds
Testing the Croissant app!
"By encoding time series as a string of numerical digits, we can frame time series forecasting as next-token prediction in text."
arxiv.org/abs/2310.07820
Code:
github.com/ngruver/llmt...
#NeurIPS 2023 Paper
Tomorrow, October 24th, 2023, I'm stoked to attend the #QuantStrats conference in London, UK 🎉
Check out the event details here: www.alphaevents.com/events-quant...
My focus? #GenerativeAI, #ReinforcementLearning, and #LOBmodelling. Can't wait!