Concepts Beyond

Job Description Must be US Citizen. Shape Safer Skies with Artificial Intelligence! Concepts Beyond is seeking a hands-on AI / Data Engineer & Analyst to serve as our AI technical lead for aviation safety and air traffic management programs. A major focus is real-time voice processing - transcribing and extracting meaning from pilot-controller radio communications - alongside anomaly detection, aircraft conflict/collision risk prediction, and predictive safety analytics. This is a high-ownership, individual-contributor role: you will personally build, train, and deploy the models and pipelines that drive real decisions in the National Airspace System. We welcome recent Ph.D. graduates, postdocs, government or military technologists, and academic engineers - sharp, fast-learning technologists ready to grow into a highly visible, public-facing leadership role. Concepts Beyond builds solutions advancing Air Traffic Management, Information Management, cybersecurity, AI, data analytics, and safety. What You'll Do Voice & Speech Processing (Priority Focus Area) Build and fine-tune real-time ASR for pilot-controller radio communications using modern architectures (Whisper-class, Conformer/FastConformer, streaming RNN-T/CTC), adapted to noisy radio channels, call-sign phraseology, and diverse accents. Implement call-sign/command spotting and Named-Entity Recognition (NER) with contextual biasing against surveillance data (e.g., ADS-B); develop speaker-role/turn detection, voice activity detection, and read-back error detection. Deploy low-latency streaming inference for live ATC operations, evaluated on call-sign and command accuracy, not just overall word error rate. AI, Machine Learning & Intelligent Systems Develop RAG pipelines over enterprise vector stores, and fine-tune and deploy LLMs/SLMs and agentic architectures for real-time operational analysis and decision support. Develop AI/ML for aviation safety analytics: aircraft conflict and collision risk prediction, trajectory-based anomaly detection, and probabilistic, time-series, and causal-factor analysis on ASIAS/FOQA/ASRS data. Data Engineering & Infrastructure Architect, build, and maintain scalable, idempotent ETL/ELT pipelines for structured and unstructured data (e.g., Apache Airflow, Prefect, Spark, Flink) across cloud and hybrid environments. Integrate and manage data across enterprise platforms (e.g., Palantir Foundry, AWS, Azure, GCP). Thought Leadership & Innovation Analyze state-of-the-art technologies and drive cutting-edge AI strategy; promote thought leadership through publications, conference presentations, and industry collaboration. Required Qualifications Ph.D., Master's, or Bachelor's in Computer Science, Engineering, or a related field. Recent Ph.D. graduates and postdocs with strong applied AI/ML or speech/NLP research are strongly encouraged to apply. 5+ years of relevant experience in data engineering or applied AI/ML; equivalent Ph.D./postdoctoral research will be considered in place of industry tenure. Hands-on experience building real-time voice/speech AI (ASR/STT, speaker/role detection, keyword or entity spotting) - the most important technical qualification for this role - plus LLMs/SLMs, RAG, and production deployment with model versioning and monitoring. Proficient in Python with strong engineering practices (testing, logging, version control); experience with ETL/ELT orchestration, SQL, and big-data frameworks (Spark, Flink) on AWS, Azure, or GCP. Fast learner and clear communicator, able to explain complex technical work to technical and non-technical audiences, and eager to grow into public speaking and industry engagement. Nice to Have Strongly preferred: aviation domain experience (ATC operations, pilot communications/phraseology, ASIAS, SWIM, FOQA/ASRS) and hands-on Palantir Foundry experience - candidates with both are an especially strong fit. Exceptional AI/ML or voice talent from adjacent regulated domains (finance, healthcare, defense) will also be considered. Voice stack: Whisper large-v3/turbo and fine-tuned variants, NVIDIA NeMo/Canary, wav2vec 2.0/HuBERT, Conformer/FastConformer, BERT-based speaker-role classifiers, Silero VAD, pyannote diarization, keyword/wake-word spotting; audio-native multimodal LLMs a plus. Probabilistic modeling and causal inference (PyMC, Pyro, Stan); vector databases (FAISS, pgvector, Pinecone); Kafka/streaming; Docker, Kubernetes, and infrastructure-as-code. DevSecOps experience in regulated or safety-critical environments; experience leading technical architecture discussions with customers and partners. LLM/SLM fine-tuning (PyTorch, Hugging Face, LoRA/QLoRA); MLOps including drift detection, retraining pipelines, and monitoring. Computer vision (OpenCV, YOLO, SAM) and multimodal models for surveillance and surface imagery.