lmeyerov

50 ms·

511.
by lmeyerov 3y ago
Brutal news, the Graphistry team would point to Noteable internally as an example of good execution on notebooks-as-a-service For folks impacted, we've been pursuing a similar-but-different product & technology vision for making da
512.
by lmeyerov 3y ago
I'm not sure why all RAG implementations would do that? Ex: Louie.AI will use a combination of tools to answer a question (database queries, RAG vector index lookups, Python, interactive charts, ...), and will often do multiple attempt
513.
by lmeyerov 3y ago
Computers that can reliably write programs / higher-order logic has long been one of my big fundamental bars for useful AGI. It will clearly get better, but for a lot of what knowledge workers already do... The bar has been cleared, a
514.
by lmeyerov 3y ago
We regularly use pygraphistry to generate /import => viz 100k+ entity embeddings on mobile fine: https://github.com/graphistry/pygraphistry , should be pretty easy to redo / scale . WebGL is impressively c
515.
by lmeyerov 3y ago
We make it pretty easy to go from networkx or any other pydata (DF, csv, parquet, ...) to interactive GPU viz w all sorts of analytics built in: https://github.com/graphistry/pygraphistry#explore-any-data-...
516.
by lmeyerov 3y ago
Ex: An agent pipeline is asked to generate some code, triggers a code agent, and azure openai is hooked up with both primary and fallback service regions + a few deployments of 3.5 turbo, 4, etc. Which model gets used first, second, ...? Wh
517.
by lmeyerov 3y ago
openai and langchain natively support all of that or expose the hooks to do it all that stuff is what we needed to do to support multiple providers vs hard-coding a single one. none of it is specific to our codebase. if we were going to tak
518.
by lmeyerov 3y ago
We've been having to build this inhouse for louie.ai to be more engineer-grade as all the libs we evaluated weren't flexible enough for common scenarios, so maybe a useful evaluation laundry list for makers & users here: * 12-
519.
by lmeyerov 3y ago
Ah you're right, LATS GPT3.5 is 84 while standalone GPT4 is 87 Given standalone GPT3.5 is "just" 48.. it's less about beating and more about meeting RE:Good Enough & Feel... very much agreed. I find it very task depe
520.
by lmeyerov 3y ago
To bring some data to a sour grapes fight: https://paperswithcode.com/sota/code-generation-on-humaneval For code generation, GPT4 is getting beat by the small prompt library LATS wrapped around GPT3.5. Given the recent
521.
by lmeyerov 3y ago
It sounds like they do it naively, which is fair for a POC Even for non-batched, you can still optimize this a bunch for fairly big models without much more work. Basically, overlap io & compute: while one layer is computing, already be
522.
by lmeyerov 3y ago
Graphistry / louie.ai | Visualization Engineer | US, global | Full-time | https://www.louie.ai , https://www.graphistry.com/careers Graphistry supercharges investigations with its GPU visual graph AI platfo
523.
by lmeyerov 3y ago
yeah i'm curious how this would stack up to vllm in a batch setting long-term, bodes well as both methodologies should be combineable, just curious for current point-in-time wrt being relevant for production use
524.
by lmeyerov 3y ago
Most LLM model use shouldn't be 'raw' but as part of a smart & iterative pipeline. Ex: * reading: If you want it to do inference over a lot of context, you'll need to do multiple inferences. If each inference is fast
525.
by lmeyerov 3y ago
Vector embedding at the app tier & orchestration / compute tier make more sense for managing the dependency than the vector DB tier for the bulk of ML/AI projects I've worked on. Round tripping through the vectordb would
526.
by lmeyerov 3y ago
... Whatever you want? Even with a vector DB, embeddings are typically BYO to beginwith - dependencies not in your DB and computed in pipelines well before it. It's handy for small apps to do in DB and have support in DB for some queri
527.
by lmeyerov 3y ago
That's great, thanks! If ever up for a chat about it, let me know -- leo@graphistry
528.
by lmeyerov 3y ago
Interesting, curious what you mean by push model? Ex: Outbound - Interactive chat sessions trigger outside push update, such as to a database for data to index, new findings & enrichments, etc Ex: Inbound - API calls trigger internal ru
529.
by lmeyerov 3y ago
Yes, we work with security analysts and other folks doing data-intensive investigations & analysis in Louie.ai (think being able to use natural language to talk to you Splunk, OpenSearch, Databricks, etc. logs/news/.., and ma
530.
by lmeyerov 3y ago
Generous severance and ensure leaving on as good terms as much as possible. Ahead of time, inventory all systems and ensure logins. We had to let go several behavior-related folks over the years, and only 1 case where we really needed them
531.
by lmeyerov 3y ago
Super cool! I'm curious if there is a quality argument to be made: imagine needing to finetune k different classifiers... Before this work, we could train a single multi-label classifier by pooling the training sets, and deploy as 1 Lo
532.
by lmeyerov 3y ago
I very much believe in this direction both as a practicing data scientist, and working daily with folks who kinda know SQL or Python but are really more experts in other stuff, like security, k8s, supply chain, etc. For a slightly different
533.
by lmeyerov 3y ago
Speaking from Graphistry experience and now louie_ai, both of which are real tech: * Initial Graphistry funding was from showing compelling 10X technology demos started in my PhD through my extended network, which got us angel & seed an
534.
by lmeyerov 3y ago
Gpt is 2023 now
535.
by lmeyerov 3y ago
We like fastify for structure, types, perf, and maintainer responsiveness. I can count how many bugs we have had. The backwards incompatibility is frustrating as most of the major changes could have been done in non-breaking ways. That make
536.
by lmeyerov 3y ago
We were looking at Kata as well, especially as an 'easier' firecracker, though I forgot why we didn't go further, and I've been curious why they seem to get little attention in practice I believe part is portability, as
537.
by lmeyerov 3y ago
Good intro. I'd be curious how they do the syscall tracing, eg, strace logs as part of CI? Funny enough, we've gone the reverse path for LLM AI-generated code sandboxing for louie.ai / Graphistry . We started with container i
538.
by lmeyerov 3y ago
I'm not sure what answer you are looking for / what alternative you are suggesting? OSM is the biggest community effort - NGO, volunteer, corporate, etc - to solve data quality in GIS. The participants do everything to improve qua
539.
by lmeyerov 3y ago
Most of these piggyback on the OSM for the data tier, which $T companies work on. Afaict the discussion is more about compute on top - tile generation, interactive renderers, ... .
540.
by lmeyerov 3y ago
We do a mix: * quick PR tweaks on someone else's branch => github code editor * data science & customer success work: jupyter notebooks (GPU) & google colab (CPU) Based on those experiences, and local dev experiences, we inv

More ›