Two design-partner slots open for 2026
Hyderabad, IN · remote-friendly
AI test automation, built to run on your infrastructure
Aura QA reads your app, agrees with you on what matters, then generates and maintains the suite. Runs in your own VPC or fully on-prem — no code, DOM, or test data crosses your network boundary.
Self-healing, shown — not claimed
run #4,182 · checkout-flow.spec · 00:41s
await page
.locator('.btn-checkout-step-3')
.click()The checkout was refactored from three steps to two. The class no longer exists. A script-based suite fails here and waits for a human.
await page .getByRole('button', { name: /place order/i }) .click()
Why this element
The step asserts "complete a purchase", not "click step 3". Aura QA re-resolved against role, accessible name, and position in the flow — then re-ran the assertion to confirm the intent still holds.
312
steps executed
7
healed automatically
2
escalated to review
0
bytes left the VPC
Run figures are illustrative pending pilot completion.
The problem
Across QA organisations, engineers report burning 40–60% of their automation time on maintenance rather than new coverage — repairing what the last release broke.
Source: industry benchmark, qaskills.sh 2026
01
Most tools patch selectors after a test breaks. The suite still fails first, and someone still triages. That is a softer landing, not a solved problem.
02
Tools generate hundreds of tests before anyone has agreed what matters. You inherit coverage you did not choose and cannot defend in a release review.
03
The category is built cloud-first. If your app touches PHI, claims, or KYC data, sending your DOM and fixtures to a vendor tenant is the end of the evaluation — regardless of how good the AI is.
How it works
Mechanism 01
Every step stores the intent it asserts, not just the selector it used. When the DOM shifts, resolution re-runs against intent and the assertion is re-verified before the step is accepted.
Step
intent: place_order
Resolve
role + a11y name
Rank
flow position
Verify
re-assert outcome
Else
escalate to human
A heal is only committed if the re-assertion passes. Anything ambiguous is escalated with the candidate diff attached — never silently accepted.
Mechanism 02
The same product, three trust boundaries. Regulated teams pick the third and nothing about the workflow changes.
Managed
Techfy cloud
runner: ours models: hosted data: transits TLS setup: same day
Hybrid
Your VPC
runner: your VPC models: your keys data: never egresses setup: ~1 week
Air-gapped
Fully on-prem
runner: your metal models: local via Ollama data: no network path setup: ~2 weeks
No telemetry, no phone-home, no vendor tenant holding your DOM. In the air-gapped topology Aura QA has no outbound network path at all — the only thing crossing your boundary is the container image you pulled.
Mechanism 03
Aura QA crawls the app and returns an Alignment Report — the flows it found, ranked by risk, with what it proposes to cover and what it will deliberately skip. You sign off. Then it generates.
Flow discovered
Risk
Proposed
Claim submission → adjudication
critical
24 tests
Member eligibility lookup
high
11 tests
Marketing footer links
low
skipped
The skipped column matters as much as the covered one. It is the artefact you take into a release review when someone asks what the suite does not protect.
Studio AI
Studio AI is the same team building custom AI systems around your workflow, with your constraints. We build it, then hand it over — your team owns the code, the models, and the runbook.
See how Studio AI works →Custom AI agents for internal ops
Multi-step operators that plan, tool-call, and verify — inside your systems.
Domain-tuned models
Fine-tuned on your corpus, evaluated against your own acceptance criteria.
AI features in your product
Shipped inside your codebase and your release process, not bolted alongside.
Eval & monitoring harnesses
For AI you already run but cannot currently measure.
Legacy automation → AI
Migrating brittle script estates without losing the coverage you have.
Build → handover.
You own it at the end.
Why believe us
Fifteen years of combined experience in the testing industry — through manual, through scripted automation, into AI. We have sat in the release reviews where a flaky suite loses the argument. That is the detail this product is built out of.
Often you should — Playwright is excellent and free. The case for Aura QA starts when maintenance, not authoring, is your bottleneck: when the suite exists and the team is spending half its week repairing it. We sit on top of that stack rather than replacing it.
Two things. Their healing is reactive — tests break first, then get patched. Ours re-resolves against stored intent and re-verifies before accepting. And they are cloud-first: if your data cannot leave your network, the evaluation ends there. Aura QA runs fully on-prem against local models.
Kickoff and first outcome locked within 2 days. Discovery — data, risks, thinnest useful slice — by day 5. A working slice you can run yourself by day 14, and an Alignment Report you can take into a release review by day 21.
We are pre-revenue and taking two design partners. That is the trade: you get the roadmap shaped around your pipeline and direct access to the engineers building it, at design-partner pricing. We would rather say that plainly than manufacture case studies.
Two design partner slots · 2026
Two teams get the roadmap built around their pipeline, direct access to the engineers, and design-partner pricing held through year one. In exchange we need a real suite, a real release cadence, and honest feedback.
30 min with the technical team.
No sales deck, no discovery form.
You qualify if
01You are in healthcare, insurance, fintech, pharma, or retail
02Your engineering team is 20 or more
03Someone will own this daily — not a side project
Design-partner pricing
Talk to us / scoped
Design-partner pricing is held through year one. On-prem deployment included. We’ll quote after we understand your suite.