Auric AI Labs is a defense technology company focused on developing AI-powered autonomous systems and data-driven vision intelligence solutions to enhance India's defense capabilities. Their mission…
Auric AI Labs is a defense technology company focused on developing AI-powered autonomous systems and data-driven vision intelligence solutions to enhance India's defense capabilities. Their mission…
Your next opportunity is in here somewhere. Sign up to explore 70,000+ startups and their open roles. No spam. No gamification. Just jobs.
70,000+
Startups
81,000+
Open Roles
4,600+
New This Week
Machine Learning Engineer
Full-timeManchester, GB
Full-time • Manchester, GB
Technical Writer
ContractMunich, DE
Contract • Munich, DE
Frontend Developer
Full-timeNew York, US
Full-time • New York, US
Software Engineer
Part-timeNew York, US
Part-time • New York, US
Product Designer
ContractNew York, US
Contract • New York, US
DevOps Engineer
ContractBerlin, DE
Contract • Berlin, DE
About This Role
We’re building a next-generation Multi-INT (Multiple Intelligence) search and reasoning platform for defense and intelligence organizations globally. We need one exceptional AI architect to own the entire AI layer of this system.
This is not an LLM wrapper role. You’ll architect frontier-level reasoning systems that fuse signals, communications, human intelligence, and electronic intelligence across millions of text documents, finding patterns that are physically impossible for human analysts to detect. You own reasoning, retrieval, indexing, and model orchestration end-to-end.
Why This Is Hard
Defense organizations accumulate decades of multi-source intelligence across SIGINT, HUMINT, ELINT, RFINT, and COMINT. Millions of unstructured text documents spanning years. Manual analysis at this scale is physically impossible.
Scale:
Detect temporal cycles, co-location patterns, and behavioral baselines across 1M+ records spanning years. An 8-month deployment rhythm buried across 2,000 reports? The system finds it. No analyst could.
Fusion without fine-tuning:
Correlate evidence across multiple intelligence disciplines, with different formats, reliability levels, and precision levels. Foundation models only. Air-gapped. No external APIs.
Messy data:
Free-text reports, structured signals analysis, agent debriefs, communication intercepts. Each with different structure, terminology, and reliability. The indexing and chunking strategy has to make all of it retrievable and fusible.
What You’ll Build
What We Need
First-principles thinkers who understand AI deeply, not as a black box, but as systems you can reason about, debug, and push beyond known limits.
Foundation model internals:
Transformer architecture, attention mechanisms, positional encodings, training dynamics. Inference optimization and quantization. You should be able to explain why a model behaves a certain way, not just observe that it does.
What this Role Is Not
Calling APIs with wrapper prompts.
Basic RAG with a single vector database.
Fine-tuning on Hugging Face datasets.
CRUD apps that happen to use LLMs.
Following tutorials.
Success Metrics
Automated pattern identification across 90%+ of entities.
Complex multi-INT query latency under 10 seconds.
Anomaly detection below 5% false positives.
System handles 10M+ documents and becomes the primary analyst tool.
95% reduction in manual analysis time.
Our Team
A small, exceptional group of researchers and engineers building at the frontier of AI for national security. The strongest technical minds we could assemble here because the problem matters and the bar is the highest they’ve encountered.
We’re setting the global standard for AI-driven intelligence analysis by building smarter architectures than anyone else in this space.
"If you’re already thinking about how you’d solve these problems, we should talk!"
Deep reasoning:
Multi-hop inference across entity relationships, temporal evidence from different INT sources, contradictory reports. Uncertainty propagation. Causal vs. correlational discrimination. Confidence calibration across sources of varying reliability.
Multi-model orchestration:
Route queries across a heterogeneous ensemble of large reasoning models, smaller specialized models, and embedding models. Query classification, decomposition, intelligent result fusion, and confidence propagation across the ensemble.
Retrieval and indexing architecture:
Multi-index retrieval spanning vector search, structured entity queries, temporal event queries, and graph traversal. Chunking strategies for heterogeneous intelligence documents. Entity extraction, relationship graph construction, temporal event indexing. Parallel retrieval, heterogeneous result fusion, confidence-weighted ranking, and iterative retrieval loops.
Reasoning systems:
Multi-hop reasoning chains with evidence tracking and uncertainty propagation. Temporal reasoning over irregular event sequences. Causal and counterfactual reasoning. Confidence calibration across source diversity, corroboration, recency, and INT-specific reliability.
Graph intelligence and pattern detection:
GNN architectures for intelligence graphs with temporal evolution and heterogeneous node types. Automated pattern mining: temporal cycles, co-location analysis, anomaly scoring, link prediction. Graph reasoning integrated with LLM analysis.
OSINT fusion and disinformation detection:
Ingest and assess open-source text intelligence. Credibility scoring, cross-verification, and integration with classified sources. Coordinated campaign detection, claim verification, source attribution.
Strategic forecasting:
Multi-scenario generation from intelligence evidence. Bayesian updating as new evidence arrives. Indicator tracking and adversarial red team analysis.
Reasoning and retrieval:
Multi-agent systems. Chain-of-thought, ReAct, Tree-of-thoughts. Advanced RAG with multi-index, iterative, confidence-weighted retrieval. Query decomposition. Uncertainty quantification.
Indexing:
Strategies for heterogeneous unstructured text. Context-preserving chunking. Entity extraction and relationship graph construction. Temporal event indexing. Vector, graph, and time-series database experience.
Graph and temporal reasoning:
GNN architectures (GCN, GAT, GraphSAGE). Heterogeneous and temporal graph networks. Link prediction, anomaly detection, symbolic + neural integration.
Implementation:
Expert-level Python. GPU memory optimization. Model serving at scale.
Strong plus:
Defense/intelligence domain knowledge. Causal inference and Bayesian methods. Publications at top venues. NLP with noisy, domain-specific text.
