lmeyerov

52 ms·

541.
by lmeyerov 3y ago
We do a ton in louie.ai bc of this: * sometimes we want an LLM with longer context, faster speed, higher quality, etc: so even in a model family, in the same job, diff model configs * we do a lot of prompt tuning for agent calls, like what
542.
by lmeyerov 3y ago
Agreed. Afaict, in practice, you setup your own casual graphs and test them. This seems very academic 1950s. Interestingly, folks are finally doing more realistic experiments in the casual equiv of arch search, and genAI is giving these eff
543.
by lmeyerov 3y ago
This is a thing I'd love to see data for - the strike price discount at time of acquisition for later-stage companies. These same companies in that megaround companies probably have stock on secondary markets, which might be a good pro
544.
by lmeyerov 3y ago
for companies raising 9 figure later-stage rounds? that's not obvious to me and relevant to this case, often the investor will do a higher valuation (artificially minting a unicorn etc) for optics/vanity reasons, which eats an add
545.
by lmeyerov 3y ago
yes, this is very much stage-specific by the time of these $100M+ rounds for $1B unicorns to hire a lot of people, the 90% discount is long gone
546.
by lmeyerov 3y ago
most of the employees came in at later rounds, so play it out. ex: They'd get say $100K in options on paper, but the pitch would be the company is high-growth, so expectation of 2X, 10X, 20X, etc over next few years. That $100K is real
547.
by lmeyerov 3y ago
That feels like damning with faint praise: we encourage Louie.ai users to only do GPT4+ level models for code gen related tasks. Even GPT4 has a lot of ways to go. Saying other models are only around 3.5 for this task isn't great. I&#x
548.
by lmeyerov 3y ago
It's worth coming up with a stronger public-facing answer We went through this last year, I think we have a public one, a private one, + our actual more 'serious' telemetry (opentelemetry, ...). For the status pages, I think
549.
by lmeyerov 3y ago
Apologies - that is likely on me specifically. We are quite lean (~12 folks), and on one end, some roles get hundreds of applicants that we are slowly working through, and on the other end, applications not following our requested flow that
550.
by lmeyerov 3y ago
Graphistry | Multiple - Senior backend, AI visualization engineer, social & product marketing, senior cleared security solutions | GLOBAL REMOTE + DC | https://www.graphistry.com/careers Graphistry supercharges investig
551.
by lmeyerov 3y ago
Mermaid for message sequence diagrams is amazing -- we recently added 'GPT for diagramming' support to Louie (conversational data & compute notebooks), and combining these rich DSLs with conversational AI & an interactive
552.
by lmeyerov 3y ago
Yeah every 3-6mo we go through a big ecosystem realignment on the Graphistry stack around CUDA, RAPIDS, torch, transformers, pyg, dgl, etc. It'd be a lot harder without docker and mamba!
553.
by lmeyerov 3y ago
We provide a GPU visual graph AI platform that gets used by folks looking at relationships & correlations in their data (think cyber, fraud, social, supply chain, genomics), and came to this: https://www.graphistry.com/g
554.
by lmeyerov 3y ago
the paper is on my queue, but yeah, these kind of class imbalance & class dependency problems are why things like RGCNs and negative sampling are so popular. It's almost the typical case in cyber, fraud, supply chain, social, etc.
555.
by lmeyerov 3y ago
ACT 1 "We're hiring remote, being in NYC or London or Peru is totally fine, geographic diversity of talent is a strength! Own your schedule!" ACT 2 "Sorry folks, we just had a downround, can't control the market. BT
556.
by lmeyerov 3y ago
For the level of copying being discussed, the scenario to compare to is when Google waymo sued Uber for copying files for hundreds of millions of dollars. They were caught copying source files & design docs. Both for software products a
557.
by lmeyerov 3y ago
If you wrote it w company resources, eg, work laptop or work internet, my understanding is that the company has a strong claim to it
558.
by lmeyerov 3y ago
Yeah as soon as we write the word 'thread' or thinking about LLM API concurrency control across many user requests, all frameworks we tried are basically a wall instead of an accelerator. For a single user demo video on Twitter or
559.
by lmeyerov 3y ago
Most prominent for a surprising amount of the vendors we looked at last year ( https://gradientflow.com/the-vector-database-index/ ): - We would have paid more for storing billion+ rows in-memory than our gov & enter
560.
by lmeyerov 3y ago
Imagine "day 2" search tasks like the user wanting to filter results for a specific time range You could force the data team to go through hoops like multimodal embeddings, multiple indexes, post-filtering... or figure it out Near
561.
by lmeyerov 3y ago
Now that opensearch has ivfpq, and I assume pgvector will get it within 6mo... the current typical case for most teams, big or small, should be handled. Regular databases are all adding it, further chiseling away at the need for dedicated v
562.
by lmeyerov 3y ago
Yeah louie.ai query generation internals feel more like classic program synthesis, where's it's more about staged compilation, iterative refinement, etc, and we use gpt3/gpt4 for each stage rather than a traditional solver. &
563.
by lmeyerov 3y ago
Agreed. Our customers on the regulated side cannot use an unexplainable UI like that by law . We take a middle ground with louie.ai of showing the database queries, data transforms, chart config, and any other decision or generation. It&#x
564.
by lmeyerov 3y ago
A lot of times you can just ask. We have a model that I see a lot of others do, even if they don't publicize. We have free, OSS, and cheap SaaS tiers fine for many of our academic users, and when a small group really wants the full ent
565.
by lmeyerov 3y ago
re:splunk (and opensearch, databricks, etc), we're already doing those with louie.ai and running early self-hosted + saas cohorts. Your questions are very much the type we'd interested in exploring with you! Feel free to signup on
566.
by lmeyerov 3y ago
Keyword search won for commodity bc semantic search was weaker and more complicated. More sophisticated orgs like Google switched to semantic/hybrid for years as both problems got addressed. Now that opensearch/elasticsearch have
567.
by lmeyerov 3y ago
I agree there's room for consulting, but as a new field, there's a lot of software currently missing for each vertical. Today, that's manual labor by consultants, but as the field matures... consultants should be doing things
568.
by lmeyerov 3y ago
Not sure if intentional: That's a super frustrating program for a traditional automatic parallelization compiler, e.g., polyhedral: * glob is an i/o operation: impure * load_file is an i/o operation: impure * data.append seri
569.
by lmeyerov 3y ago
With Louie.ai, there is a lot of work on specialization for the job, and I expect the same for others. We help with data analysis, so connecting enterprise & common data sources & DBs, hooking up data tools (GPU visuals, integrated
570.
by lmeyerov 3y ago
i was curious b/c we're building a lot of this inhouse for louie.ai just out of need using the current seems unclear for us: * we need to own the data & database, and align with our regular+vector infra -- where do they live h

More ›