Founding AI/ML Engineer

Company Description

DeepTerra transforms existing fibre-optic cables into an AI-powered real-time monitoring system — no new hardware infrastructure needed. Using Distributed Acoustic Sensing (DAS), we turn dormant telecom fibre into thousands of virtual sensors that detect intrusion, theft, structural damage, and leaks across critical infrastructure — at a fraction of the cost of traditional monitoring.

We’re a team of engineers and scientists (4+ years of DAS R&D with Universiti Malaya, 2x patents pending) building toward pilot deployments with major telecom and O&G players in Southeast Asia. We’re now looking for a founding AI/ML & Data Engineer to build and continuously improve the intelligence layer of our product — DeepDAS.

The role:

DAS data is not your typical ML dataset — it’s continuous, high-volume, low-SNR waveform data, with rare true events buried in hours of noise and environmental interference (especially in Southeast Asian tropical/monsoon conditions, which is core to our data moat). You’ll own the problem of turning that raw signal into reliable, real-time detection and classification.

Specifically, you’ll:

• Build and maintain data pipelines to ingest, clean, label, and manage large volumes of DAS waveform data from field deployments and lab testing.
• Develop and improve AI/ML models for anomaly detection and event classification (e.g. intrusion, theft, leaks, structural anomalies), working closely with our Chief Scientist and hardware team to understand the physical signal characteristics behind the data.
• Design a continuous retraining loop — as we collect more field data, models should keep improving, with proper versioning, evaluation, and drift monitoring rather than one-off static training.
• Tackle real-world data challenges — sparse/noisy labels, class imbalance (rare events vs. long stretches of background noise), and environmental noise specific to tropical climates.
• Work with our software engineering team to get models deployed efficiently for real-time inference on edge hardware at customer sites.

What we are looking for:

• Direct experience working with large-scale time-series or waveform sensor data — seismic, DAS, acoustic/vibration monitoring, sonar/radar, or similarly structured continuous signal data. This is the single most important qualifying experience — general ML/AI backgrounds without time-series signal exposure are not a strong fit.
• Strong foundations in signal processing (FFT, filtering, denoising) — you don’t need to be a DSP specialist, but you need to be fluent in this language, since it’s inseparable from working with DAS data.
• Practical ML engineering experience building models that go beyond a Jupyter notebook — data pipelines, model versioning, retraining workflows, and performance monitoring in a real deployed system.
• Comfortable with messy, real-world data — sparse labels, imbalanced classes, and the need to build labeling/annotation processes largely from scratch.
• 3-8+ years of relevant experience , with a track record of taking models from research/prototype into a working, evolving production system.
• Startup mindset comfortable with ambiguity, moving fast, and working hands-on with a small, technically deep founding team.

Nice to have: experience with seismology, geophysics, oil & gas sensing, industrial IoT, or defense/security sensing systems; experience handling multi-site or multi-hardware data at scale; exposure to edge AI deployment (ONNX, TensorRT, or similar).

To apply:

Send a short note on why this excites you, along with your resume/LinkedIn/GitHub (and any relevant work with time-series or sensor data), to [email protected] — or reach out directly.

We’re moving fast and evaluating fit over formal process — expect a real conversation, not a lengthy screening funnel.