Data Science with a focus on GenAI, Large Language Models, and Vector Databases.

Data Scientist

Ford Motor Company

- Data Analytics: Leveraged GCP BigQuery, Teradata, and SQL Server to gather, understand, and analyze
quality and warranty metrics for usage in predictive modeling.

- Architecture: Developed reference architecture models, documentation, and recommendations to drive 
technology adoption including for Large Language Models, and Knowledge Graphs.

- Project Management: Lead pilot to evaluate Neo4j Aura SaaS as fully managed knowledge graph solution.

Enterprise Architect

- Software Interface: Research and develop Chatbots and Digital Assistants with many AI platforms including 
both IBM Watson and Microsoft Azure AI suites, as well as experiment with latest NLP/NLU models.

- Architecture: Advise architecture team on research initiatives to help architects design the most cutting-edge 
and implementable architectural layout for Chatbots, Digital Assistants, and Knowledge Graphs.

- Hardware Experimentation: Analyze newly released hardware such as the NVIDIA Jetson Nano for AI 
capability with open-source software such as TensorFlow and PyTorch, including image recognition.

- Leadership & Training: Lead and manage dozens of Proofs of Concept projects relating to AI & Cognitive 
Computing, and host training sessions, a Community of Interest, and symposiums on emerging technologies.

- ASR: Trained Speech-to-Text model to perform with a 4% WER on Ford specific vocabulary on noisy plant 
floor environment which intermittently reached 90dB.

- Knowledge Graph: Spearheaded the effort to bring Graph Databases into the company by running PoCs 
with business units throughout Ford including Manufacturing, legal, procurement, quality, PD, etc.

- Manage Supplier/Business Relationships: Lead and managed relationships with suppliers and business 
units throughout the enterprise to ensure emerging technologies continued to grow at Ford.

AI & Cognitive Computing Engineer

-   Secured Windows and Linux operating systems by developing a testing framework for 260 OS 
     security policies using Chef and Inspec. 

-   Designed a deployment pipeline to check for OS security vulnerabilities by spinning up ephemeral 
    VMs and running automated tests to verify compliance in HPC environments.

-   Provisioned Docker containers to increase efficiency of the security pipeline, more than tripling the 
     speed of each test.

Tools & Automation Engineer

Meijer

-Software Interface:  Used PowerShell scripting to automate tasks in System Center Configuration Manager (SCCM), VMware AirWatch, and Microsoft Distributed Transaction Coordinator (MSDTC).

-Monitoring logistics:  Created dashboards using Microsoft Visio to allow efficient viewing of server state, health, and performance information within System Center Operation Manager (SCOM).

-Agile Technique:  Employed Agile Technique to first provide Minimum Viable Product (MVP) then incrementally improve software to create a more robust solution.

ITS: Architecture and Server Infrastructure Intern

Halla Visteon Climate Control (Hanon Systems)

-Software Interface: Used Finite Element Analysis (FEA) pre/post processing software, Altair  Hypermesh and MSC Apex, to prepare CAD models for FEA by creating a robust mesh of elements.

-Software logistics: Advised managers on new program to create time-savings and increase efficiency.

-Six Sigma Green Belt: Achieved Six Sigma Green Belt certification by scripting custom commands in Hypermesh to reduce mouse travel and decrease number of clicks for large functions.

-Customer Interface:  Interacted with company MSC about a new program that could increase HVCC's productivity and savings as well as make a more lean product development process.

CAE Intern

Lowe's Companies, Inc.

Product logistics: Stocked product, signed off on incoming product
Customer Interface: Loaded customer's cars with heavy product
Safety: Spotted for drivers of heavy machines such as forklifts

Loader

Product logistics: Stock and rotate product 
Customer Interface: Sold meat and seafood to customers
Safety:  Kept area to federal sanitary guidelines

Meat Department Clerk

Data Science @ KX.

I'm Ryan – your go-to enthusiast for all things generative AI, vector databases, and knowledge graphs. My passion is to contribute to this rapidly evolving space by engaging with these emerging technologies, sharing my findings, and absorbing knowledge from this community.

I specialize in developing and leading technical and architectural projects with technologies such as Retrieval Augmented Generation (RAG) and tuning the interplay of LLMs and Vector DBs. I have worked in the world of enterprise chatbots, natural language processing (NLP), and training custom acoustic and linguistic automatic speech recognition (ASR) models.

Connect with me and let’s exchange insights, brainstorm solutions, or just have a chat about how these technologies are shaping our lives.

The first open-source evaluation harness for LLMs on q just dropped 🤯 .

Why does it matter?

 👉 Python HumanEval scores are peaking around 96%, but on q-HumanEval, top models are still at 43.4% Pass@1 = Big opportunity.

What’s packed in this release:
 🧠 Standardized LLM benchmarks for q
 📊 Public leaderboard for fair comparisons
 🧪 164 q-HumanEval tasks (diverse + real-world)
 🔁 Python benchmarks can now test q solutions (!)

Who should care?
 👩‍💻 Devs: Find the best models for real-world q tasks
 🛠️ LLM builders: Track and tune performance on q-specific tasks

Early results?
 💥 Grok 4 outperforms Claude 4 on q
 🔁 Multi-iteration loops (CodeAct style) boost Grok from 43.37% → 74.32% Pass@10 — wild!

What’s next?
 🚧 q-MBPP benchmarks, agentic tracks, new metrics
 🤖 Not just model output — real agentic behavior coming to the leaderboard

Want in?
Clone the repo, try it out, contribute some tasks — let’s push LLMs for q forward, together.

#qLang #kdbPlus #LLMs #OpenSource #Benchmarking #AgenticAI #Grok #KX #AI4Finance

Ryan Siegler

Developer Advocate | GenAI | Vector DBs | Data Science

ML/RL Research Engineer ​> Post-training ​> Reasoning | Ex-founder

🚀 Introducing the q Evaluation Harness: The First Open-Source Evaluation Framework for LLMs on q/kdb+

Why it's crucial:
🔍 Top models hit ~96% Pass@1 on Python HumanEval…
⚡ But the best q-HumanEval score? Just 43.4%.
 That’s a huge performance gap and exactly what we’re here to close.

What it brings to the table
✅ A standardized way to measure LLM performance on the q language
✅ Easy extension to other Python datasets/benchmarks, since existing Python tests can now be used to validate q solutions
✅ A q-HumanEval benchmark featuring 164 diverse tasks
✅ A transparent public leaderboard that supports fair and open performance comparisons

Who benefits:
👩‍💻 Developers: Identify top-performing models and select best tools/models for real-world q implementations.
🧪 LLM builders: Use a standard benchmark to evaluate LLMs performance on q, track improvements over time, and guide training toward the areas that matter most for real-world q applications.

Early results:
🏆 Grok 4 shines in q-HumanEval, outperforming the widely adopted Claude 4 in general programming.
🔁 Leveraging multiple iterations substantially enhances scores (e.g., Grok 4: 43.37% → 74.32% Pass@10).
💡 This shows that prioritizing agentic, CodeAct-style loops yields significant improvements.

What comes next:
📈 Add new datasets, including q-MBPP, along with new metrics
🧠 Introduce an agentic track on the leaderboard to evaluate AI agents on q, not just raw model outputs

Get involved:
🛠 Try it out, run experiments & share your findings
 📥 Contribute new tasks & help shape the next wave of q benchmarks

📂 Repository: https://lnkd.in/g-Zni54B
📖 Learn more: https://lnkd.in/gsNv2Su6
🥇 Leaderboard: https://lnkd.in/gQqCNi2Q

Special thanks to Andrew Morrison for his invaluable contributions to this project and Kamala Parvathanathan for leading it.

& Thanks, Nataraj Dasgupta, for connecting everyone and making this project happen. 

Built by KX

Open article: Introducing q Evaluation Harness: The First Open-Source Evaluation Framework for LLMs on q/kdb+ by medium.com

Introducing q Evaluation Harness: The First Open-Source Evaluation Framework for LLMs on q/kdb+

Taplio

Ryan Siegler's Linkedin Analytics

's Best Posts (last 30 days)

Want to drive more opportunities from LinkedIn?

Taplio

.css-1jgreyo{background-image:linear-gradient(284deg, rgba(39,43,72,1) 0%, rgba(33,47,183,1) 100%);color:transparent;-webkit-background-clip:text;background-clip:text;}Ryan Siegler's Linkedin Analytics

's Best Posts .css-qqfgvy{font-size:var(--chakra-fontSizes-md);}(last 30 days)

Want to drive more opportunities from LinkedIn?

Ryan Siegler's Linkedin Analytics

's Best Posts (last 30 days)