The migratory patterns of Arctic terns span nearly the entire globe, a journey unmatched by any other living creature
QyveraPTE — practice, scored, tracked
AI scores all four skills on every attempt — mocks roll into one climbing band score.
ProductSoftware research & engineering lab
Our work lives somewhere between research and reality. We follow ideas wherever they lead, sometimes into products of our own, sometimes into collaborations with founders pursuing ideas worth believing in.
Featured works
The migratory patterns of Arctic terns span nearly the entire globe, a journey unmatched by any other living creature
AI scores all four skills on every attempt — mocks roll into one climbing band score.
ProductAbstract—We present a framework for optimizing the deployment of cloud-native workloads across heterogeneous providers and regions. Placement is cast as a multi-objective problem over cost, latency and reliability, and solved to select configurations that satisfy service-level objectives (SLOs) while minimizing spend. On representative workloads it lowers cost by up to 31% at equal or better tail latency.
Index Terms—cloud computing, deployment optimization, multi-objective optimization, SLO, latency, reliability.
Modern applications are increasingly deployed across multiple cloud providers and geographic regions. Choosing where to place each service is a delicate trade-off between operating cost, end-to-end latency and fault tolerance. Static, hand-tuned configurations rarely remain optimal as traffic and provider pricing shift over hours and days.
We model deployment as a constrained optimization over a set of candidate placements. Let x denote an assignment of services to regions and instance classes, annotated with measured price pi and per-service latency:
where L(x) is the end-to-end tail latency and A(x) the composed availability of the chosen placement, both differentiable in the relaxed assignment.
A guided branch-and-bound searches the space using cost and latency lower bounds for pruning, warm-starting from the previous solution to make re-optimization cheap under drift.
Across three workloads and four providers, the framework lowers cost by 18–31% while meeting all SLOs, and adapts within seconds to price and traffic changes.
Treating deployment as continuous multi-objective optimization yields cheaper, more reliable placements than static design.
Deploy cheaper without breaking SLOs — up to 31% saved.
Research · IEEEOne app for the whole clan — a living family tree, shared causes & a transparent fund.
Community App9 PRs merged — SIMD search, HNSW & embedding infra.
Open sourceSpeed cameras that can't lie — plates read at the edge, every violation anchored on Solana.
Web3 / BlockchainEvery mission has a beginning. Ours started with a conversation about problems worth solving.
The more we explored those ideas, the more one thing became clear: we didn’t just want to build software, we wanted to build the kind of engineering lab we wished existed. A place where research leads the way, curiosity is expected, and every product is treated like a craft.
Today, we’re a small team with different strengths but one shared belief: obsession beats everything. We challenge each other, learn relentlessly, listen closely to the people we build for, and keep refining until the work speaks for itself.
We’re here for the love of the game.
And this is just the beginning!

Believes the best ideas are shaped through curiosity, thoughtful execution, and a willingness to keep improving.

Values genuine relationships above all else, creating opportunities through trust, openness, and meaningful conversations.

Research-driven product thinker with a sharp eye for detail and a passion for solving real problems.

Software wizard who never went to university, yet mastered everything from computer fundamentals to modern frameworks.

An intersection of technology and business, shaping products with both perspectives in mind.
Get in touch
Tell us about your product, project, or research — a web app, platform, AI product, or anything else — and we’ll get back within one business day.
hello@qyvera.com