We build intelligent tools that scan the sky, detect anomalies, and surface discoveries -- autonomously, at scale.
Local-first. Cloud-ready. Built for researchers and production.
Products
Intelligent tools for automated astronomical discovery, from optical image analysis to radio signal processing.
v1.1.0
AI-powered astronomical image anomaly detection
v0.1
High-performance radio signal anomaly detection
Purpose
Modern astronomical surveys generate 20 terabytes of data per night. The next generation of telescopes (LSST/Rubin Observatory) will produce even more. The bottleneck in astronomy has shifted from collecting observations to analysing them. Most survey images are classified once by a standard pipeline and never examined again. Unusual objects — galaxy mergers, transient events, gravitational lenses — slip through unnoticed.
We built AstroLens and AstroSETI because this problem is solvable with modern ML engineering. An ensemble of specialised models (Vision Transformers for classification, out-of-distribution detection for anomalies, YOLO for transient localisation) can systematically screen every image and flag what deserves a second look — without human fatigue, bias, or time constraints.
These tools demonstrate what we deliver professionally: end-to-end ML systems that move from a research concept to a production-grade, self-operating pipeline. The same engineering approach — designing for autonomy, building self-correcting feedback loops, instrumenting for observability — applies to any domain where organisations need to extract signal from large-scale data streams.
Autonomous anomaly detection that runs for days with zero intervention, self-corrects, and improves its own models during operation.
Validated against real astronomical data. Independently recovered known published objects (supernovae, galaxy mergers, gravitational lenses) without prior knowledge.
The architecture patterns — streaming ML pipelines, self-correcting thresholds, autonomous operation — apply to any domain that processes high-volume data at scale.
Results
Autonomous analysis across multiple sky surveys with zero human intervention.
Technology
An end-to-end intelligent pipeline from raw sky survey data to actionable discovery reports.
Runs on laptop: CPU, Apple MPS, and NVIDIA CUDA supported
Infrastructure
Universal solution patterns that take ML and AI workloads from local research environments into production-ready cloud deployments.
Current open-source tools run locally on any laptop. This is the research and experimentation layer.
Production deployment on AWS. Managed services orchestrate the full ML pipeline at scale.
End-to-end delivery covering design, implementation, automation, and operations. Full lifecycle ownership from solution architecture through to production reliability.
Terraform, CloudFormation, CDK -- repeatable, version-controlled infrastructure
GitHub Actions, CodePipeline -- automated build, test, and blue-green deployment to ECS/EKS
Declarative deployments, drift detection, and rollback from a single source of truth
ECS and EKS workload management, auto-scaling, service mesh, multi-AZ resilience
IAM policies, GuardDuty, Security Hub, automated vulnerability scanning in pipeline
CloudWatch, X-Ray, Prometheus, Grafana -- full-stack monitoring and distributed tracing
Right-sizing, Savings Plans, Spot instances, Cost Explorer dashboards
About
Technical Director -- Cloud, DevOps & Platform Engineering
Saman is a specialist in AWS Cloud and DevOps with 10+ years of experience delivering cloud platform consultancy and production-grade solutions for large organisations. He designs and automates secure, scalable AWS architectures end-to-end—covering solution design, CI/CD, platform engineering, and technical leadership—with a focus on building reusable patterns that teams can ship, operate, and evolve with confidence.
Before specialising in cloud, Saman built a strong software engineering foundation across Linux embedded systems, monitoring applications, and cross-platform desktop development using Qt, including low-level networking and protocol implementation in C. That mix of deep engineering roots and modern AWS delivery means he brings practical judgement to every engagement—balancing speed, reliability, security, and cost without the hype.
Philosophy
Science advances fastest when tools are open, reproducible, and accessible to everyone.
Transparent and reproducible research that anyone can verify, audit, and build upon.
No cloud dependency required. Run everything on your own hardware, with your own data.
Designed to plug into existing research pipelines with standard interfaces and formats.
Community-driven development and improvement. Contributions, feedback, and collaboration welcome.
Get in Touch
Open to research collaborations, industry partnerships, and commercial engagements.
Partner on astronomical surveys, anomaly detection research, or radio signal analysis projects.
Deploy AI-powered astronomical analysis at scale with cloud-native architecture and support.
Custom ML pipelines, data processing systems, and infrastructure tailored to your research needs.