We build software,
then we run it.
SoftStackers is a studio building a portfolio of AI-native products. Software we design, ship, and operate ourselves. Different markets, one operating engine: small senior teams, AI at the core, built to last on our own infrastructure.
HaloSilver
Runs quietly on a parent's or loved one's computer, reads what's on screen for scam signals with AI, and sends the family a calm heads-up: romance grooming, gift-card pressure, fake tech-support, remote-access. You see the alert, never their screen.
SonicStacker
Tell it the story (the dog's name, the inside joke) and it writes and produces an original song in about ninety seconds. Birthdays, weddings, memorials, just because. Every track original, never trained on protected works.
PolicyChanges
Sweeps every payer (UnitedHealthcare, Aetna, CMS, BCBS and more) every 24 hours, then uses AI to surface each change with its impact level, affected billing codes, effective dates, and the exact action a practice needs to take. Before it hits revenue.
Ask about the products, the AI work, or working with us.
It answers from what we've actually built, and says so when it doesn't know. Leave an email and it passes you to the team.
CLAUDE ON AMAZON BEDROCK · SCOPED TO SOFTSTACKERS
AI-native, not AI-added
Every product starts with the model at the center: agents, retrieval, and tuned models doing real work. Not a chat box bolted onto old software.
We operate what we ship
We don't hand off and walk away. We run our products in production on our own AWS infrastructure, and keep improving them for years.
Small, senior, fast
No layers, no handoffs. The people who design a product are the people who build and run it. That's how a whole portfolio comes from one small team.
The AI work behind the products.
We are not consultants writing about AI. Every capability below is running in production in something you can go use today.
Agentic systems
Multi-step agents that plan, call tools, and recover when a step fails. Guardrails, retries, and human handoff where the stakes are real.
Retrieval (RAG)
Hybrid search, deliberate chunking, reranking, and grounded answers with citations. Tuned against real user queries, not a demo corpus.
Tiered model routing
A cheap high-recall model screens every candidate; a stronger judge model rules on what survives. Most of the quality at a fraction of the inference cost.
Evals and observability
Golden datasets, offline eval suites, and regression gates before release. Every production model call traced and reviewable.
Workflow automation
AI that removes whole steps from an operation rather than sitting beside it. Measured by work eliminated, not messages sent.
Vision and audio
On-screen understanding and generative audio running on real hardware — including a desktop agent light enough for a ten-year-old PC.
A product company with years of shipping behind it.
SoftStackers has built and operated production software since 2019. Founded by Ben Rodrigue, we spent years shipping and running systems for other companies. That engineering depth now goes into a portfolio of our own AI-native products.
We also take on a small number of applied-AI engagements each year for select partners, and remain an AWS Select Tier Partner. Talk to us →
Builders who want to own the whole thing.
We're a small senior team shipping real products to real users. If you want range (design, build, and operate AI-native software end to end) we're always talking to exceptional people.
Using a product, hiring us, or looking to invest?
Product support lives on each product's own site. For partnerships, press, careers, or investment, reach the studio directly.
contact@softstackers.com