spitfire
I'm a data scientist with more than 15 years experience who has worked with major automakers, shipping companies, and F100 companies, I'm looking for my next project. An ideal fit is a small/medium company with green or brownfield problems that need solving.
I'm skilled at building from 0 to 1, I normally do what is now called data engineer, data scientist, and machine learning engineer as one person. My work usually involves finding non-traditional datasources - Once it was USDA data for rural community demographic enrichment, Several times it was getting hands on with people or equipment. If we work together I'll do what I need to in order to solve the problem.
Technologies/techniques: Machine learning/AI, LLM's, All the tools surrounding ML and AI, A whole host of others.
Contact me if you have a tough challenge I can gnaw on.
I'm a forward-deployed engineer. My background is 15+ years of data science, but I focus on solving problems. I work in messy, ambiguous, regulated environments. Often at a client's site, inside their systems and their data. I excavate unmapped data, learn how the team works, and ship tools that stay in production.
Deployed and still running: Legal clause extraction and validation; site rust detection and management; routing and predictive-maintenance systems that stayed in service.
I compress engineering, data, and modeling into one function, owned end to end. I show up, figure out what needs to be built, and make it work.
Tooling: ML/AI, LLMs, and the engineering around them: evals, determinism, observability, guardrails. The tech is just a tool, The hard part is knowing what to build.
If you have a messy, ambiguous problem that needs someone who can embed and ship, my email is in my profile.
I'm a data scientist with more than 15 years experience who has worked with major automakers, shipping companies, and F100 companies, I'm looking for my next project. An ideal fit is a small/medium company with green or brownfield problems that need solving.
I'm skilled at building from 0 to 1, I normally do what is now called data engineer, data scientist, and machine learning engineer as one person. My work usually involves finding non-traditional datasources - Once it was USDA data for rural community demographic enrichment, Several times it was getting hands on with people or equipment. If we work together I'll do what I need to in order to solve the problem.
Technologies/techniques: Machine learning/AI, LLM's/RAG, All the tools surrounding ML and AI, A whole host of others.
Contact me if you have a tough challenge I can gnaw on.
Location: West coast, remote only.
I'm a data scientist with more than 15 years experience who has worked with major automakers, shipping companies, and F100 companies, I'm looking for my next project. An ideal fit is a small/medium company with green or brownfield problems that need solving.
I'm skilled at building from 0 to 1, I normally do what is now called data engineer, data scientist, and machine learning engineer as one person. My work usually involves finding non-traditional datasources - Once it was USDA data for rural community demographic enrichment, Several times it was getting hands on with people or equipment. If we work together I'll do what I need to in order to solve the problem.
Technologies/techniques: Machine learning/AI, LLM's/RAG, All the tools surrounding ML and AI, A whole host of others.
Contact me if you have a tough challenge I can gnaw on.
Location: West coast Canada, remote only.
Techniques: Agentic & Trajectory development, Machine learning/AI, LLM's/RAG, Transformers, GenAI, MCMC, MCTS, Bayesnets, Neural architecture search.
Languages: Python, Postgres, Spark, Bash, Mathematica, Others.
Frameworks: PyData (Pandas, NumPy, scikit-learn), PyTorch, TensorFlow/Keras, PySpark, LangChain, XGBoost, LightGBM, Others.
Tools: AWS, Spark, Gurobi/highs/ortools, Kubernetes, Docker, Linux, Git, Tableau/databricks
I'm a data scientist who has worked with the likes of Porsche and Mulesoft, I'm looking for the next thing.
Some past work:
- In the wild part failure prediction for automotive parts.
- Server fleet failure prediction for Mulesoft. Giving a warning to do load shedding.
- Real-time Vehicle routing, Get a couch from where you are, to where you want it to be.
- Taxi fleet maintenance optimization and failure prediction for a latam taxi company. Saved entire cars.
- Automated sports highlights - Feed in video, get nicely cut highlights of diving catches, touchdowns, etc.
- Wound identification and mangement - Identify what sort of wound, label and categorize the tissue.
- Account grading and revenue optimization for ARM.
- Maritime anti-piracy - Identifying pirate skiffs, at risk vessels, and routing them aroound danger.
Fantastic.
You could build fast software today by simply adopting a reference platform, say. A 10 year old 4core system. then measuring performance there. If it lags then do whatever work needs to be done to speed it up.
Personally I think we should all adopt the raspberry pi zero as a reference platform.
Edit: * office 2000 was fast too with 32 megs of ram. Seriously what have we done?
If I ever retire and run a bar on the beach^w^w^w^w retro computer store, that’s what it’ll be called.
Outsized success a curse.
My understanding is that they had the Czech business location with Russian developers so they had a clean public face.
I’d believe you if you insisted on Kiev based too, I don’t know anything first hand.
I’m not sure your second point is true. The vast vast majority of classified information is very boring, or operational like frequencies of radar, etc.
Both sides know the basics, it’s what frequencies the radar comms and aircraft work at that is classified.
There’s very little “OMG this one algorithm changes everything!!”. Unless proven otherwise
I’m not aware of any computer science breakthroughs required for the f35.
Move a few flags around in a few registers and for all practical purposes it’s stuck.
I expect a crash program to reengineer them has already started if only unofficially.
I would love to spend the time and see how altering the query alters the reasoning path. How firm is in the path once it’s chosen?
A high level approach has the possibility to be very computer efficient.
So once they go down a path they can’t properly backtrack.
This feels like the ground truth I’ve experienced in LLMs to date.
But then you look at it and think Yeah, obviously they're not going to have MTBF times in the millions. It's going to be hundreds of hours - once a week, or maybe every few weeks between real hard crashes.
How would that change your behaviour.
For exactly the reasons you mention. They’ll fit right in.
It is available on annas archive. Along with other similar books (Search: "Planning in the USSR").
LOOK UPON ME AND DESPAIR!
Obviously it didn’t work out though.
The other book worth getting is “red plenty”.
Location: West coast, remote only.
I'm a data scientist with over 20 years experience. I've worked with the likes of Porsche and Mulesoft, I'm looking for the next thing.
An ideal fit would be a small/medium company with hair on fire green or brownfield problems that need solving. I'm currently quite interested in AI, and routing/optimization problems.
Technologies/techniques: Machine learning/AI, LLM's/RAG, Transformers, GenAI, MCMC, MCTS, Bayesnets, Neural architecture search, Python, Pytorch/Tensorflow/Keras, All the tools surrounding ML and AI, A whole host of others over a 25 year career.
Some past work: - In the wild part failure prediction for Porsche, saving millions. - Server fleet failure prediction for Mulesoft. Give a 5 minute warning to do load shedding. - LLM system to output human readable flight delay information for well known airline. - Real-time Vehicle routing, Get a couch from where you are, to where you want it to be. - Goal oriented voice agent in the pre-LLM era. - Taxi fleet maintenance optimization and failure prediction for a latam taxi company. Saved entire cars. - Automated sports highlights - Feed in video, get nicely cut highlights of diving catches, touchdowns, etc. - Wound identification and mangement - Identify what sort of wound, label and categorize the tissue. - Account grading and revenue optimization for ARM. - Maritime anti-piracy - Identifying pirate skiffs, at risk vessels, and routing them aroound danger.
Who shouldn't contact me:
- Ads or gambling. I'd rather sleep at night. Thanks.Location: West coast, remote only.
I'm a data scientist with over 20 years experience. I've worked with the likes of Porsche and Mulesoft, I'm looking for the next thing.
An ideal fit would be a small/medium company with hair on fire green or brownfield problems that need solving. I'm currently quite interested in AI, and routing/optimization problems.
Technologies/techniques: Machine learning/AI, LLM's/RAG, Transformers, GenAI, MCMC, MCTS, Bayesnets, Neural architecture search, Python, Pytorch/Tensorflow/Keras, All the tools surrounding ML and AI, A whole host of others over a 25 year career.
Some past work: - In the wild part failure prediction for Porsche, saving millions. - Server fleet failure prediction for Mulesoft. Give a 5 minute warning to do load shedding. - LLM system to output human readable flight delay information for well known airline. - Real-time Vehicle routing, Get a couch from where you are, to where you want it to be. - Goal oriented voice agent in the pre-LLM era. - Taxi fleet maintenance optimization and failure prediction for a latam taxi company. Saved entire cars. - Automated sports highlights - Feed in video, get nicely cut highlights of diving catches, touchdowns, etc. - Wound identification and mangement - Identify what sort of wound, label and categorize the tissue. - Account grading and revenue optimization for ARM. - Maritime anti-piracy - Identifying pirate skiffs, at risk vessels, and routing them aroound danger.
Who shouldn't contact me:
- Ads or gambling. I'd rather sleep at night. Thanks.