The best AI products aren't the smartest. They're the ones that change safely, every week.
That is what we build: the models, workflows, and infrastructure that let a team ship fast without shipping regressions. Measured outcomes from this quarter's client work in consumer finance, enterprise AI, and consumer media.
01 / Consumer finance
+40% click-through, and still climbing in production
For a lending and collections company, we built a model that sends each borrower's message at their most responsive moment instead of on a fixed schedule. Click-through rose 40% over a randomized baseline, and keeps outperforming it live.
Payment integrations that took weeks now ship in days
For a lending and collections platform, we turned integration-building from a bespoke effort into a repeatable system: reusable AI skills that scaffold each new payment processor, review the code, and keep task tracking current on their own. Three new processor integrations shipped in August alone, and every future one inherits the same machinery, with hooks already in place to add AI acceleration wherever it pays.
02 / Enterprise AI
AI that surfaces the risks nobody thought to ask about
For an enterprise AI platform, we built a context graph that turns fragmented signals across CRM, calls, emails, and conversations into persistent, connected organizational memory. Identities resolve across sources, timelines persist, and the system reasons about what happened and what should have happened, so a prolonged silence or a shifting stakeholder stops going unnoticed. The AI moves from summarizing data to proactively flagging hidden pipeline risk and emerging opportunities, with far less time spent piecing context together by hand.
An enterprise agent ~25% faster, on ~40% fewer premium tokens
For an enterprise AI agent, we treated context as a strategic part of the workflow instead of handing the model everything it might need. The system now retrieves only the structures and signals relevant to the specific business question before passing the task to the reasoning model. The agent runs ~25% faster on ~40% fewer premium-model tokens, with no cut to the depth of analysis. Faster, cheaper, and more reliable, from context engineering alone.
200k lines of Java rebuilt as 50k of Python, in two weeks
A client's production engine had become the bottleneck. Its dev environment went down almost every day for two to three months, and a mobile app launch slipped three or four times waiting on it. We rebuilt it from the ground up, verified against the real clients, and cut over in two weeks with zero rollbacks and no downtime. A quarter of the code, and the daily outages stopped.
03 / Consumer media
$1–2 cheaper and 30–75 seconds faster on every render
For a consumer AI music-video product, we removed a paid lip-sync step by carrying the song straight into generation. Each render now costs less, finishes sooner, and has one fewer thing that can fail.
80–90% more assets per creator, ~140 steps down to one
For a creative marketplace, bulk intake and reusable assets replaced a twenty-form, 120-upload catalogue process with a single submission, and let creators reuse what they make instead of rebuilding it each time.
New video and image models judged in days, not a rebuild cycle
We built an evaluation studio for generative video and image models, nine of them across four vendors, so identity drift, distorted motion, and quality regressions are caught on reproducible evidence before customers ever see them.
04 / About Unravel
We ship to production every week. Our own products come out of the same AI-first workflows we build for clients.
StenoMeeting notes that write themselves, clean and searchable when the call ends.steno.unravel.tech → thingShip your coding agents' output as live, commentable, versioned pages. This letter runs on it.usething.ai → TabloKeeps a human in the loop on your AI coding agents, live from the corner of the screen.tablo.unravel.tech → AsklightCtrl+Space, ask anything. AI one keystroke away, with free models and nothing to set up.asklight.unravel.tech →Planning an AI product, or a rebuild that can't afford downtime?
Thirty minutes, and we'll tell you straight what it takes, what it's worth, and how fast we can do it.
Book a call or reply to this email — it reaches us directly at kapil@unravel.tech