Ichi | Remote (US) | Full-Time | Senior Software Engineer, Full Stack | LLM-powered construction code compliance

Ichi is building an AI-powered professional toolset to transform construction permitting and code compliance. We're helping municipalities, plan examiners, architects, and construction professionals streamline the permitting process to reduce bottlenecks—ultimately accelerating construction timelines in an $11T global industry.

We're a small team (you == employee #6) backed by Costanoa Ventures looking for a high-ownership Senior Software Engineer (8+ years exp) who can build from scratch and deeply cares about product quality. You'll design and implement product features that integrate intuitive UX and powerful multi-modal (and multi-model) LLM pipelines, working directly with users to refine our approach.

Tech stack: Ruby, Rails, Python, React, TypeScript, GraphQL, Postgres, AWS.

Must have: Full stack experience, product intuition, and interest in LLMs.

CS degree or equivalent preferred; we work on hard problems and first-principles thinking and attention to detail are essential.

Our name "Ichi" (Japanese for "city", and "one") reflects our mission to improve the physical spaces around us through never-before-possible technology and bridging a fundamental information asymmetry.

Apply: https://www.ichiplan.com/senior-software-engineer

One of apple’s biggest missed with “AI” in my opinion, is not building a universal search.

For all the hype LLM generation gets, I think the rise of LLM-backed “semantic” embedding search does not get enough attention. It’s used in RAG (which inherits the hallucinatory problems), but seems underutilized elsewhere.

The worst (and coincidentally/paradoxically I use the most) searches I’ve seen is Gmail and Dropbox, both of which cannot find emails or files that I know exist, even if using the exact email subject and file name keywords.

Apple could arguably solve this with a universal search SDK, and I’d value this far more than yet-another-summarize-this-paragraph tool.

I have this same issue with gmail. I can not find e-mails by an exact word from text or subject. It is there, but search would not show it. I don't understand how a number one email provider can fail at that.

Or, say, a number one search provider ;)

and search provider! Of all the companies in the world, why is Gmail search just not better?

For this to happen they'd have to actually pay attention to spotlight and the quicklook/spotlight plugin ecosystem they abandoned. There's lots of obvious ways to combine LLMs with macOS' unique software advantages (applescript, bundle extensibility) but they have spent years systematically burning those bridges. I don't think they'll be able to swallow their pride and walk everything back.

I run engineering for a venture backed AI-first startup and we use Ruby/Rails.

For us, it made sense to leverage one of the best domain modeling and ORM frameworks out there. Most of our inference is http calls to foundational models, but we can still fine tune and host models on GPUs using Python.

Inference matters, but part of building an effective user platform are the same old SaaS problems we’ve had before, and Rails just works. Inbound and outbound email done in a day. Turning an OCR’d title from ALL CAPS into Title Case is one method call and not a whole custom algorithm, etc.

A lot of people seem to think Ruby is slow for some reason but it’s as fast as Python, and with falcon as fast as node for async behavior. Safe to say the application language taking 0.03 seconds instead of 0.003 seconds when you have to wait 3 seconds for first token is absolutely not the bottleneck with LLM heavy workflows, anyway.

And yes, metaprogramming is a powerful tool with which you can easily shoot yourself in the foot. We culturally just don’t write any code that’s not greppable so don’t use method_missing kinds of things unless it’s in a robust gem like active record. Pretty trivial problem to solve really.

PS - We’re hiring, if that philosophy aligns with you!

What's a good way to learn the modern Ruby ecosystem nowadays?

I played with Ruby when I was a teenager (~2015 or so), and I definitely remember enjoying it. I know there's still a vocal group of users who love it, so I would be interested in digging in again.

Yea, just agreeing with the other commenter. Creating a Rails app is the way to go to get started with Ruby. https://guides.rubyonrails.org/getting_started.html

It's my favorite programming language but I seldom get to use it because I'm an AI Engineer. But I just recently went out on my own so I guess that can change now, hm...

I would actually start with the Rails Guides docs, they’re very good and running the given commands should actually work:

https://guides.rubyonrails.org/getting_started.htm

Just have a toy app you want to build in mind

Rails still the “batteries included” piece of Ruby. Recently added parts like Hotwire, SolidQueue etc are pretty interesting to know.

Outside of it, you might find interesting libraries like sinatra, sequel, roda, dryrb, faraday, sorbet, truffle ruby…

In terms of LLM code generation as well, the well structured nature of a Rails application, where there is a place for everything, a structure for tests to be added, really helps from the perspective of getting a comprehensible application out of it that is easy to modify. In addition to the existence of well tested component for most normal web application tasks, maybe it helps that a lot of Rails has already been based on old-fashioned code generation for 20 years.

I have this same suspicion. I dusted off a hobby Rails app from two years ago I was making with Cursor. I decided to try completely changing the main functionality of the app with the much better LLMs of today and was shocked how well it did with one-shot.

Now compare that to my recent experience with having Cursor help me work on a preexisting Node/React app...geez. What a pain. (It doesn't help that I wasn't the one that originally created the React app though.)

Absolutely love this approach. The power and versatility Rails provides for data modeling is top notch. You're 100% right that many of the problems you must solve when building a SaaS app are the same.

> Finally, building HNSW indices in Postgres is still extremely slow (even with parallel index builds), so it is difficult to experiment with index hyperparameters at scale

For anyone coming across this without much experience here, for building these indexes in pgvector it makes a massive difference to increase your maintenance memory above the default. Either as a separate db like whakim mentioned, or for specific maintenance periods depending on your use case.

``` SHOW maintenance_work_mem; SET maintenance_work_mem = X; ```

In one of our semantic search use cases, we control the ingestion of the searchable content (laws, basically) so we can control when and how we choose to index it. And then I've set up classic relational db indexing (in addition to vector indexing) for our quite predictable query patterns.

For us that means our actual semantic db query takes about 10ms.

Starting from 10s of millions of entries, filtered to ~50k (jurisdictionally, in our case) relevant ones and then performing vector similarity search with topK/limit.

Built into our ORM and zero round-trip latency to Pinecone or syncing issues.

EDIT: I imagine whakim has more experience than me and YMMV, just sharing lesson learned. Even with higher maintenance mem the index building is super slow for HNSW

Thanks for sharing! Yes, getting good performance out of pgvector with even a trivial amount of data requires a bit of understanding of how Postgres works.