Senior AI/ML Engineer

Los Angeles

Engineering – Artificial Intelligence / Full-Time / On-Site

The Opportunity We are building a next-generation AI intelligence platform that ingests data from satellite feeds, autonomous sensors, logistics networks, and open-source streams — fusing it all into a live knowledge graph and producing calibrated probabilistic assessments in real time. This is not another chatbot or RAG wrapper. This is serious, adversarially-aware ML infrastructure where the cost of a wrong prediction is asymmetric and real. We're looking for the engineer who builds the ML brain — someone who thinks in probability distributions, treats data poisoning as a first-class design constraint, and ships production systems that operate in high-assurance, compliance-driven environments. What You'll Build:

Multi-Model Prediction Engine- Own the full lifecycle of a suite of specialist prediction models forming the platform's inference core:

Temporal event prediction — time-to-event modeling using Weibull survival analysis and Bayesian posterior updating over multi-domain precursor signals
Activity convergence modeling — detecting logistics and movement patterns indicative of coordinated operational activity
Supply chain & logistics risk forecasting — disruption and depletion modeling across tracked supply networks
Behavioral attribution — fingerprinting and pattern-of-life modeling for entity identification and operator attribution
Trajectory prediction — kinematic and intent-driven modeling for autonomous system and vehicle movement
Composite threat scoring — multi-signal index construction with Platt-calibrated confidence intervals
Effectiveness degradation modeling — attrition estimation incorporating observational inputs
Long-horizon behavioral baseline construction and anomaly detection over tracked entities

Bayesian Inference & Calibration:

Implement Bayesian posterior updating pipelines across hot, warm, and cold ingestion tiers with distinct latency targets
Ensure all model outputs are Platt-calibrated with explicit confidence intervals suitable for operational decision-making
Design and operate a multi-model fusion layer correlating specialist outputs into tiered alerts and composite assessments
Architect ensemble mechanisms that preserve model independence and prevent correlated failure modes

Data Fusion & Knowledge Graph Engineering:

Own a multi-stage ingestion pipeline across video/imagery streams, logistics feeds, structured intelligence data, satellite/orbital feeds, and OSINT
Maintain and extend a domain ontology across Neo4j (graph), Qdrant (vector store), and Apache Iceberg (columnar analytical storage)
Design entity resolution, deduplication, and confidence-weighted merge strategies for multi-source entity upserts
Implement temporal versioning of knowledge graph state to support replay, audit, and counterfactual analysis

Adversarial Pattern Detection:

Implement DBSCAN spatiotemporal clustering for event aggregation across multi-domain signal streams
Apply Dynamic Time Warping (DTW) for pattern matching against historical operational sequences
Design and operate deception detection algorithms that identify adversarial signal manipulation and data poisoning attempts
Develop adversarial robustness testing protocols for all pattern detection components

MLOps & Inference Serving:

Deploy and operate NVIDIA Triton Inference Server for all specialist models
Build model versioning, A/B evaluation, and zero-downtime promotion pipelines
Implement human-in-the-loop feedback collection interfaces feeding back into model improvement cycles
Build replayable, auditable training lineage records compliant with enterprise security and export control requirements
Establish model performance KPIs: calibration error, intervention frequency, prediction lead time, false alert rate

What You Bring:

Bachelor's or Master's in Computer Science, Statistics, Applied Mathematics, or related engineering field
5+ years hands-on building and operating production ML systems — not notebooks, shipped systems
Solid working knowledge of Bayesian inference, Weibull survival analysis, and probabilistic calibration (Platt scaling, isotonic regression)
Production experience with Qdrant, Neo4j, and Apache Iceberg (or equivalent columnar storage)
Hands-on experience with NVIDIA Triton Inference Server or equivalent high-throughput model serving
Strong Python engineering skills with production-grade standards; PostgreSQL/pgvector and GCP cloud-native experience
Experience building spatiotemporal event aggregation pipelines and anomaly detection over streaming multi-source data
Familiarity with MCP (Model Context Protocol) server architecture or equivalent tool-layer orchestration patterns

Preferred:

Experience with data interchange and messaging compliance standards in deployed operational systems
Familiarity with CesiumJS or equivalent 3D geospatial visualization frameworks
Experience building adversarial robustness testing frameworks for ML systems in contested data environments
TypeScript proficiency for orchestration tool endpoint development

Key Competencies: 1- Probabilistic Rigor Builds systems where uncertainty is measured, calibrated, and communicated explicitly — no hiding behind point estimates 2- Adversarial Awareness Designs for active deception, data poisoning, and signal manipulation as first-class threat vectors, not edge cases 3- Operational Judgment Understands the asymmetric cost of false negatives vs. false positives in high-stakes prediction contexts 4- Systems Depth Reasons from raw pipeline latency through model architecture and inference serving to agent behavior and analyst-facing outputs 5- Standards Fluency Works within compliance frameworks and export control requirements as engineering constraints, not administrative overhead This role is for builders who want their work to matter at scale. If you've shipped ML systems into real operational environments and you're ready to work on problems with genuine consequence — we'd like to talk.

Apply for this job