Rails is one part of the practice. Levelbrook also ships Python work directly: FastAPI and Django backends, data pipelines and automation, and AI/LLM tooling. The proof is deployed and public below. Same senior engineer, same clean corp-to-corp terms.
A visual history pipeline for heavy-equipment rental: 20,000+ photos across 40 machines, embedded with DINOv2, reduced with UMAP, clustered with HDBSCAN, and auto-labelled zero-shot with CLIP — then served through a React viewer. A working, deployed Python ML system, not a notebook.
# 20k+ rental photos -> a browsable visual
# history, clustered by machine & viewpoint.
import torch, umap
from sklearn.cluster import HDBSCAN
feats = dinov2.encode(photos) # ViT-B/14
coords = umap.UMAP(n_neighbors=15).fit_transform(feats)
labels = HDBSCAN(min_cluster_size=12).fit_predict(coords)
# zero-shot names for each cluster via CLIP
names = clip.zero_shot(centroids(coords, labels),
prompts=EQUIPMENT_VIEWS)Levelbrook is not a one-framework shop. A lot of valuable contract engineering is better done in Python: a data pipeline, an automation that removes a manual process, a FastAPI service, or the glue around a large language model. Python is a first-class part of what we offer, alongside Rails and Node.
The credibility for that isn't a list of years — it's a shipped, deployed system you can click on right now. equipment-cluster (featured above) is a real Python ML pipeline: 20,000+ heavy-equipment rental photos embedded with DINOv2, reduced with UMAP, clustered with HDBSCAN, auto-labelled zero-shot with CLIP, and served through a React viewer. That's PyTorch, scikit-learn, and modern vision models wired into something that actually runs.
Python engagements run on the same clean footing as the Rails work: corp-to-corp through Levelbrook LLC, with MSA, SOW, mutual NDA, and a COI (GL + Professional Liability, $1M / $1M) ready on day one. Work ships as real PRs with written-down decisions, and you keep maintainable code — not a dependency on the contractor.
FastAPI / Django / Flask services and REST/GraphQL APIs — typed, tested, and documented for the consumers who'll actually call them.
ETL pipelines, scraping, report generation, and the internal tooling that removes a recurring manual task from someone's week.
RAG, agents, embeddings, and vector search — plus the retries, evals, and guardrails that move an LLM feature from demo to production.
Wiring vision and embedding models into a working system — exactly the shape of the deployed equipment-cluster project above.
Levelbrook is a senior software consulting practice across Rails, Python, and Node. Python is backed by a shipped, deployed ML project (equipment-cluster) you can try above, not just a claim.
FastAPI and Django most often for services and APIs, Flask where it's lighter-weight, plus the data/ML stack: PyTorch, scikit-learn, UMAP, open_clip, pandas, and the usual vector-search tooling.
Yes — RAG pipelines, agent workflows, embeddings and vector search, and the evaluation and reliability plumbing around them. The equipment-cluster project shows the applied-ML side of that.
Identically to the Rails work: corp-to-corp through Levelbrook LLC, with negotiable rates for ongoing engagements, fixed-scope projects quoted per engagement. MSA / SOW / NDA / COI ready on day one.
Often that's the point — a Rails app with a Python data or ML service alongside it. One accountable senior engineer across both keeps the seams clean.
Send the brief. You'll get an honest read on whether it's a fit and how we'd scope it within one business day.