“I hired Harrison in 2016 at Kensho and that was the most impactful thing I did for the firm in my 3 years there. He is willing to do whatever it takes to help the team, from feature engineering/experiment running, to tricky backend tasks that he had no idea how to do and figured out. For example, he figured out how to cut down the memory overhead of our multiprocessing workloads by 5x just by reading about how multiprocessing works and profiling the code, most data scientists I have worked with do not go that deep into "backend" programming. He is also very strong at prioritizing next steps and isolating bottlenecks. For example, he identified a moment when a model was good enough to put in front of users (I wanted to keep making it better) and reoriented me towards doing the data engineering necessary to get the model out the door. His next step intuition is so good that even when I was meant to be managing him I would send him my plan for every day to get his feedback, and this ended up saving me a huge amount of time. Harrison is a very strong python programmer and a careful code reviewer — he will take the bugs that nobody wants to fix, even if he didn’t create them, or provide a useful solution concept to a less experienced programmer if he thinks they would benefit from implementing it themselves. Finally, Harrison has an uncanny ability to generate significant model improvements by looking model at errors and thinking of features that might explain those differences. This skill is not only relevant with Timeseries and text data, which we worked on at Kensho — we did a sports project together and it was exactly the same. In conclusion, Harrison knows exactly what features need to be built and consistently gets to a point where the model is good enough to productionize, and also has the skills to productionize and get feedback from users.”
San Francisco, California, United States
Contact Info
15K followers
500+ connections
About
Activity
-
💡 Few-shot prompting to improve tool-calling performance I'm very bullish on few-shot prompting, but there aren't a ton of resources on best…
💡 Few-shot prompting to improve tool-calling performance I'm very bullish on few-shot prompting, but there aren't a ton of resources on best…
Shared by Harrison Chase
-
���What is the future of planning for agents? Yes - model improvements will help models get better at planning. But we think good prompting and…
🔮What is the future of planning for agents? Yes - model improvements will help models get better at planning. But we think good prompting and…
Shared by Harrison Chase
Experience & Education
Recommendations received
1 person has recommended Harrison
Join now to viewMore activity by Harrison
-
🛃Generic vs custom cognitive architectures To improve the planning ability of an agent, you may try to improve the cognitive architecture. You can…
🛃Generic vs custom cognitive architectures To improve the planning ability of an agent, you may try to improve the cognitive architecture. You can…
Shared by Harrison Chase
-
💻GPT Computer Assistant This popular OSS project (4.6⭐️) is an alternative work for providing ChatGPT MacOS app to Windows and Linux Great example…
💻GPT Computer Assistant This popular OSS project (4.6⭐️) is an alternative work for providing ChatGPT MacOS app to Windows and Linux Great example…
Liked by Harrison Chase
-
🧠Planning for agents One of the big limitations for agents working reliably is their ability to plan. I talk about that in a new blog post. Covers:…
🧠Planning for agents One of the big limitations for agents working reliably is their ability to plan. I talk about that in a new blog post. Covers:…
Shared by Harrison Chase
-
This time on my journey to make cool stuff, I had the opportunity to beta test LangGraph Cloud- LangChain’s one stop solution for deploying graph…
This time on my journey to make cool stuff, I had the opportunity to beta test LangGraph Cloud- LangChain’s one stop solution for deploying graph…
Liked by Harrison Chase
-
The new LangChain-Qdrant integration with sparse/hybrid retrieval is now available! 🚀 Key features: ✔ Support for dense, sparse, and hybrid vector…
The new LangChain-Qdrant integration with sparse/hybrid retrieval is now available! 🚀 Key features: ✔ Support for dense, sparse, and hybrid vector…
Liked by Harrison Chase
-
This is a fun project we made to show off some of the "time-travel" aspects of LangGraph checkpointing You can iteratively write a story in a…
This is a fun project we made to show off some of the "time-travel" aspects of LangGraph checkpointing You can iteratively write a story in a…
Shared by Harrison Chase
-
✍️ Choose-your-own-adventure story writing with LLMs ✍️ In the video below, we build an app that can do story writing in a choose-your-own-adventure…
✍️ Choose-your-own-adventure story writing with LLMs ✍️ In the video below, we build an app that can do story writing in a choose-your-own-adventure…
Liked by Harrison Chase
-
95% of tutorials online just show you how to create a simple conversational chatbot, but in the real-world is very different. You need to maintain…
95% of tutorials online just show you how to create a simple conversational chatbot, but in the real-world is very different. You need to maintain…
Liked by Harrison Chase
Other similar profiles
Explore collaborative articles
We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.
Explore MoreOthers named Harrison Chase in United States
-
Harrison Woodard
Finance Major at James Madison University | CIS Minor | European Business Concentration
-
Harrison Chase
-
Harrison Chase
Attending New Mexico Institute of Mining and Technology
-
Harrison Chase
Product Manager at Studio NoCode
10 others named Harrison Chase in United States are on LinkedIn
See others named Harrison Chase