Two products. Both live. Both production-grade.
DeepfieldLabs builds cloud-native data products for problems where the answer has to be defensible. DebrisGuard (primary) re-scores 30,000+ catalogued satellites every cycle and ships insurer-verifiable, HMAC-signed orbital-anomaly alerts. MitraSETI Cloud accelerates radio-telescope signal analysis 45× for SETI researchers.
Every alert ships with its inputs, model version, and a public verification URL. That is the same evidence chain a regulator or underwriter would demand, built in from day one rather than assembled after the fact.
Mission Control retains the document for 2 years, so it still verifies when a claim hits adjusters in 2028.
LeoLabs and Slingshot start at $50 k+/year, sales-led, 6-week cycle. DebrisGuard starts at $0 (Watcher) → $299/mo (Analyst), self-serve, on the same public-catalog inputs.
Cost gap: 99 %. Capability gap on what most operators actually need: ~0. pip install debrisguard-bench and re-verify any alert end-to-end.
Two products, both live
DeepfieldLabs builds and operates both, on the same engineering chassis. DebrisGuard is the commercial flagship, designed for satellite operators, brokers and insurers who need defensible alerts. MitraSETI Cloud is the research product, designed for SETI groups and university labs who need turboSETI without the laptop. Both have free tiers and both are production-grade.
Primary product · Live
Every cycle we re-score 30,896 catalogued objects from Space-Track and Celestrak. Three independent detectors (catalog d², BSTAR-drift, and manoeuvre fingerprinting) fuse into a single trustable anomaly score, and every alert ships with an HMAC-signed forensic dossier the recipient can re-verify with one command.
The BSTAR-drift early-decay detector reaches ROC-AUC = 0.91 (95 % CI 0.88–0.92) at a 45-day horizon against a 522-decay Space-Track backtest. In the last 14 days the manoeuvre engine flagged 1,667 burns across 821 active satellites, with 74.6 % confirmed by post-burn TLE convergence. The free tier is the real product, not a demo.
ROC-AUC = 0.91 [95 % CI 0.88–0.92] at the 45-day horizon on n_pos = 522 actually-decayed Space-Track objects. Bootstrap CIs are significant above chance; harness re-runs publicly on demand.
1,667 burns flagged in the last 14-day window across 821 manoeuvring objects, with 74.6 % confirmed by post-burn TLE convergence. Each carries an escalate / review / monitor chip, and the rate is published, not asserted.
Every alert backed by an HMAC-signed JSON document with TLE provenance, scoring trace, and reproducibility metadata. Court-ready.
World-map view with subpoint trails, alert stream, conjunction badges, decay-watch panel, and pipeline-health KPIs. Refreshes every page load.
GET /v1/objects, /v1/reports/daily/latest, /v1/manoeuvres/recent, /v1/fleet. Slack/Teams webhooks on Operator tier. Cognito JWT or API key auth.
Atomic DDB counters prevent over-quota writes even at 50-thread concurrency. HTTP 429 with retry_after + tier-aware upgrade link.
Research product · Live
The same engineering toolkit, applied to radio-telescope data. A Rust-accelerated SETI engine runs 45× faster than turboSETI, a CNN + Transformer classifier separates signals across 9 categories, and results export as FITS catalogs ready for downstream pipelines.
The free tier processes 10 .h5 / .fil files a month, enough for verification, hobbyist analysis, or building on top of the REST API. Pricing and tier details live on the MitraSETI Cloud site, not here.
Compiled Rust engine. Files that took hours locally finish in under a minute.
CNN + Transformer models for narrowband, drifting, broadband, RFI, and 5 other classes.
Results include signal-persistence tracking across observations; ready for downstream pipelines.
Why DeepfieldLabs
DebrisGuard and MitraSETI Cloud are the proof. The same engineering chassis is available for custom projects, built around your data, your compliance requirements, and your users.
Need your platform in Frankfurt for EU compliance? Sydney for APAC latency? Virginia for cost? We deploy in any cloud region. Your users get local performance, you keep full control over where your infrastructure lives.
Your data stays in the region you select. We enforce strict boundaries so files never leave the jurisdiction you choose. Essential for regulated industries, government work, and international research collaborations.
Every organisation has different needs. We design cloud architectures tailored to your workload: containers, Kubernetes, serverless, or a hybrid approach. You tell us what you need, we figure out the best way to build it.
Components can span multiple cloud providers. Data might live in one cloud while compute runs in another. We work across AWS, GCP, and Azure to design the right topology for performance, cost, and compliance.
Infrastructure as code, automated deployments, monitoring, and security hardening. What we ship is what runs in production, built to handle real traffic from the start.
Every product starts with a free tier. As usage grows, upgrade with no migration and no downtime. Your data and configuration carry forward automatically.
We design and deploy custom cloud platforms across AWS, Google Cloud, and Azure. Your workloads run where they make the most sense for performance, cost, and compliance.
Primary compute, storage, and ML infrastructure
ML workloads, analytics, and data processing
Enterprise integrations and hybrid workloads
Managed clusters across EKS, GKE, AKS with unified orchestration
ECS Fargate and Cloud Run for zero-ops container workloads
Event-driven Lambda, Functions, and step workflows at scale
Model training, inference, and pipelines on SageMaker & Vertex AI
Region-locked storage ensures data never leaves your chosen jurisdiction. Full compliance with GDPR, HIPAA, and institutional regulations.
Region-lockedEdge caching, intelligent DNS routing, and multi-region failover. Users experience local performance no matter where they are.
Multi-regionEnd-to-end encryption, WAF, IAM, secrets management, and automated compliance scanning at every infrastructure layer.
End-to-endOpen Source
The core analysis engines behind our cloud products are open source. Use them locally, or let the cloud handle the infrastructure.
Rust-accelerated SETI signal analysis engine. 45x faster than turboSETI. CNN + Transformer classifier. The same engine powering MitraSETI Cloud.
Documentation & Source →
AI-powered astronomical image anomaly detection. Vision Transformers + YOLO. 22,195 images analysed with 99.5% accuracy.
Documentation & Source →Get in Touch
Questions about the platform, enterprise deployments, custom solutions, or research collaboration? We'd love to hear from you.
DeepfieldLabs