We're building an agentic data analytics companion for intralogistics.

Looking for a hands-on founder/CTO who can design and ship an agent-based system end-to-end. You're comfortable building an agent harness, write clean typed Python, and have experience with tools like uv, DuckDB, and PydanticAI. You must be based in Germany.

If this sounds like you, email: michele.dallachiesa+logenta-cto2@sigforge.com with subject "HN Who's Hiring - Logenta CTO". Include your CV and answer one question: "why me?".

We're building an agentic data analytics companion for intralogistics.

Looking for a hands-on founder/CTO who can design and ship an agent-based system end-to-end. You're comfortable building an agent harness, write clean typed Python, and have experience with tools like uv, DuckDB, and PydanticAI. You should be based in Europe (Germany preferred, and Hamburg is best).

If this sounds like you, email: michele.dallachiesa+logenta-cto@sigforge.com with subject "HN Who's Hiring - Logenta CTO". Include your CV and answer one question: "why me?".

EDIT: Unfortunately, if you are not already based in the EU, we won't be able to proceed.

I specialize in exploratory data, ML and AI projects. Past clients include Google, NASA, and the UK, HK governments on topics such as infrastructure, cybersecurity, robotics, and decentralized finance. If you're facing high-stakes challenges or need insights from messy data, let’s connect.

LINKEDIN: https://www.linkedin.com/in/dallachiesa/

EMAIL: michele.dallachiesa@sigforge.com

I specialize in exploratory data/ML/AI projects, working with clients like Google, NASA, and the UK/HK governments across infrastructure, cybersecurity, robotics, and decentralized finance. If you're facing high-stakes challenges or need insights from messy data, let’s connect.

LINKEDIN: https://www.linkedin.com/in/dallachiesa/

EMAIL: michele.dallachiesa@sigforge.com

OFFERING: With two decades of experience developing analytical and predictive tools, I am good at explorative projects.

WEBSITE: https://www.sigforge.com/

LINKEDIN: https://www.linkedin.com/in/dallachiesa/

EMAIL: michele.dallachiesa@sigforge.com

OFFERING: Algorithms, crypto, finance, hardware, and AI. Past recent projects on forecasting for infrastructure projects, on-chain zero-knowledge proofs, LLMs for NL2SQL, and AI for mechatronics. Just fire me an email at michele.dallachiesa@sigforge.com

There is no verifiable information, seemingly from China; they have a massive count of stars/forks. My attempts to contact a few of them failed, so I tend to think these are bots.

What's their purpose?

Comments URL: https://news.ycombinator.com/item?id=40308500

Points: 5

# Comments: 1

Comments URL: https://news.ycombinator.com/item?id=40003071

Points: 2

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=39932074

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=39608985

Points: 1

# Comments: 0

Comments URL: https://news.ycombinator.com/item?id=39517046

Points: 1

# Comments: 0

3) “Can it also track the model's state during training if, e.g., there is an early stop, and then I want to continue the training process?”

With MLtraq, You can dump and load arbitrary objects, including model weights and other state parameters. Let's consider the example https://mltraq.com/howto/02-artifacts-storage/. MLtraq dumps and reloads from the filesystem the binary blobs referenced in the tracked metadata. Similarly, you can store artifacts in third-party services and data stores.

Today, I’m unveiling MLtraq (https://mltraq.com), an open-source Python library for AI developers to design, execute and share experiments. It allows you to track anything, reproduce, collaborate, and resume the computation state anywhere.

KEY BENEFITS

- Extreme Tracking and Interoperability: With native database types, native serialization in Numpy and PyArrow, and a safe subset of opcodes for Python pickles, unprecedented (and safe) tracking capabilities.

- Promoting Collaboration: Work seamlessly with your team by creating, storing, reloading, mixing, resuming, and sharing experiments using any local or cloud SQL database.

- Flexible: Interact with your experiments using Python, Pandas, and SQL from Python scripts, Jupyter notebooks, and dashboards without vendor lock-in.

DOCUMENTATION AND CODE

- Documentation: https://www.mltraq.com

- Source code: https://github.com/elehcimd/mltraq

Thoughts? Looking forward to your feedback. I hope you enjoy it! Thank You! (You can also reach out to me directly; my email is on my profile page.)

Cheers, Michele

PS. Sharing a few questions I have addressed so far:

1) “How does it differ from MLflow”?

There’s an overlap in features, but the scope is different:

With MLtraq, tracking the state is so transparent that it feels like checkpointing the experiment for later analysis and continuation. With robust/flexible serialization, experiments can easily copy/load to new databases. In MLflow, tracking is designed to track metrics and only a little more. The setup is less flexible, but you have more readily available integrations.

With MLflow, the emphasis is on covering the complete lifecycle, including model versioning and artifacts storage. MLtraq emphasizes experimentation, with an excellent model for experiments inspired by state monads from functional programming that encourages incapsulation/composition and parameter grids to simplify exploration.

In summary, MLflow is a better fit if you prioritize MLOps. MLtraq is a good candidate for experimentation. --

2) “Does it work for Torch models, too?”

Let’s start with the example “IRIS Flowers Classification” at https://mltraq.com/#example-3-iris-flowers-classification. Using https://skorch.readthedocs.io/en/stable/classifier.html, we can add one more scikit-learn compatible model. Alternatively, one can redesign the train_predict step without using scikit-learn. The results will include the accuracy score for the newly added model.

Comments URL: https://news.ycombinator.com/item?id=39274088

Points: 2

# Comments: 1

Together with Oxford Global Projects, we have built a family of forecasting models for S-curves using data from a total of 2,700 years of combined construction activity, with an aggregate cash flow of USD 60bn.

The S-curve in project management is a graphical representation that illustrates the cumulative progress of a project over time. It is called an "S-curve" because its characteristic shape resembles the letter "S": It starts slowly, accelerates, and then levels off.

Project delays and budget overruns are often linked with anomalies within the expenditure profile, like a sluggish burn rate or unexpectedly high spending towards anticipated project completion. Timely identification of these anomalies empowers proactive intervention to realign projects on the path to success.

This short video shows how we're modelling expenditure curves to enable many use cases, including spending projections, cost overruns and underruns, outlier analysis, and more.

Comments URL: https://news.ycombinator.com/item?id=38516178

Points: 1

# Comments: 1

OFFERING: Data Products & AI Consulting (Freelance). I have a strong track record in building reliable data pipelines, dashboards and end-to-end AI solutions. CV and references are available upon request.

RECENT PROJECTS: Empowering Researchers with Personalized Recommendations, Accelerating Public Consultations with Large Language Models (LLMs), and Guarding High-Risk Large Infrastructure Projects with an Early Warning System.

TECHNOLOGY STACK: Data Science/AI: Pandas, Polars, NumPy, JupyterLab, Matplotlib, Scikit-learn, PyTorch, Hugging Face Transformers; Data engineering/BI: PostgreSQL, Spark, Snowflake, Dask, Joblib, Airflow, Celery, Fabric, Docker, FastAPI, Alembic, AWS (EC2, EMR, S3, Lambda, Cloud- Watch), GCP (AI Platform, Compute Engine, Storage, Dataform), Looker, Metabase.

CONTACT: michele.dallachiesa@sigforge.com | https://www.linkedin.com/in/dallachiesa

Which repositories show the fastest growth? Are there any notable patterns worth highlighting? Let’s find out! The analysis considers 158 public GitHub code repositories at OpenAI, Anthropic and Cohere, created since August 2014, with an aggregate of 12k commits.

The S-curve in project management is a graphical representation that illustrates the cumulative progress of a project over time. It is called an "S-curve" because its characteristic shape resembles the letter "S": It starts slowly, accelerates, and then levels off.

The cumulative number of code commits over time can be used as raw data to model development progress “cost” with S-curves. Similar results can be obtained with the count of distinct authors (harder to control) and the count of modified files.

The animation illustrates the progression of commits over time, with normalisation applied to both axes. Each frame captures a snapshot of the repositories at a specific moment in time. A combination of the fastest and slowest repositories is highlighted with colors and labels. Quiet projects cluster in the top-left corner, and accelerating projects are found in the bottom-right area.

Over time, patterns tend to stabilise. Projects with synchronised acceleration can be attributed to coordinated commits from private repositories. The Python APIs for OpenAI, Anthropic, and Cohere stand out as some of the most active repositories, with OpenAI taking a prominent role in the Node.js / Typescript API development. 5 out of 8 of the most active repositories belong to OpenAI.

S-curves in software development are well-equipped to run simulations, comparative performance analyses, identify project delays and anomalies, and optimise resources in large teams with multiple projects.

Comments URL: https://news.ycombinator.com/item?id=38390564

Points: 4

# Comments: 1

OFFERING: Data Products & AI Consulting (Freelance). I have a strong track record in building reliable data pipelines, dashboards and end-to-end AI solutions. CV and references are available upon request.

RECENT PROJECTS: Empowering Researchers with Personalized Recommendations, Accelerating Public Consultations with Large Language Models (LLMs), and Guarding High-Risk Large Infrastructure Projects with an Early Warning System.

TECHNOLOGY STACK: Data Science/AI: Pandas, NumPy, JupyterLab, Matplotlib, Scikit-learn, PyTorch, Hugging Face Transformers; Data engineering/BI: PostgreSQL, Spark, Snowflake, Dask, Joblib, Airflow, Celery, Fabric, Docker, FastAPI, Alembic, AWS (EC2, EMR, S3, Lambda, Cloud- Watch), GCP (AI Platform, Compute Engine, Storage, Dataform), Looker, Metabase.

CONTACT: michele.dallachiesa@sigforge.com | https://www.linkedin.com/in/dallachiesa

Data Products & AI Consulting (Freelance). I work with clients in US and EU on projects lasting 1-12 months. CV and references available upon request.

PAST PROJECTS: Predicting demand for contact center services; Determining the effectiveness of marketing campaigns; Outdoor advertising; Natural language processing; Making predictions and categorizing data using statistical models; Improving traffic flow in urban areas.

TECHNOLOGY STACK: Data Science/AI: Pandas, NumPy, JupyterLab, Matplotlib, Scikit-learn, PyTorch, Hugging Face Transformers; Data engineering/BI: PostgreSQL, Spark, Snowflake, Dask, Joblib, Airflow, Celery, Fabric, Docker, FastAPI, Alembic, AWS (EC2, EMR, S3, Lambda, Cloud- Watch), GCP (AI Platform, Compute Engine, Storage, Dataform), Looker, Metabase.

CONTACT: Email: michele.dallachiesa@sigforge.com; LinkedIn: https://www.linkedin.com/in/dallachiesa