Neural Language Processing Platform | TeraSystemsAI Careers

Platform Overview

Domain-Specific Language Understanding

Our NLP Platform delivers custom transformer architectures fine-tuned for specialized domains where general-purpose models fail. Through rigorous domain adaptation, synthetic data generation, and active learning workflows, we achieve state-of-the-art performance on legal contract analysis, medical record extraction, financial document understanding, and technical patent search.

Built for enterprises requiring interpretable AI with privacy guarantees, our platform provides explicit confidence scores, attention visualization, and traceable reasoning chains. Every prediction includes uncertainty quantification and model provenance, enabling auditable decision-making in regulated industries. On-premise deployment options ensure sensitive data never leaves your infrastructure.

Transformer Architectures

Custom BERT, RoBERTa, and T5 variants for domain-specific tasks
Long-context models handling 16K+ token documents
Efficient attention mechanisms (Longformer, BigBird, Reformer)
Multi-task learning across NER, classification, QA, summarization
Continual learning and model updating without catastrophic forgetting

Domain Adaptation

Legal corpus pre-training on 10M+ contracts, case law, regulations
Medical NLP trained on MIMIC-III, PubMed, clinical guidelines
Financial document understanding for 10-K, earnings calls, SEC filings
Technical patent analysis and prior art search optimization
Active learning workflows reducing labeling costs by 80%

Interpretability and Trust

Attention visualization showing model reasoning process
SHAP and LIME explanations for individual predictions
Confidence calibration with uncertainty quantification
Adversarial robustness testing against input perturbations
Model cards documenting performance, limitations, intended use

Multilingual Support

50+ languages including low-resource languages
Cross-lingual transfer learning for zero-shot tasks
mBERT and XLM-R based architectures for multilingual understanding
Language detection and code-switching handling
Translation quality estimation and MT post-editing

Privacy and Security

On-premise deployment with air-gapped model serving
Federated learning training without centralizing sensitive data
Differential privacy guarantees for model training and inference
PII redaction and de-identification pipelines
SOC2 Type II and HIPAA-compliant infrastructure

Production Infrastructure

RESTful API with batch processing and streaming endpoints
Kubernetes-based autoscaling handling 10B+ requests/month
GPU-optimized inference with TensorRT and ONNX Runtime
Model versioning, A/B testing, and canary deployments
Comprehensive monitoring, alerting, and SLA guarantees

Join the NLP Platform Team

Build state-of-the-art language models solving real-world problems in legal, medical, and financial domains

Hiring planned → Coming soon

Principal Research Scientist, NLP

Location: Philadelphia, PA or Remote

Compensation: $220,000 - $310,000 + equity

Type: Full-time

Lead NLP research agenda advancing state-of-the-art in domain-specific language understanding for legal, medical, and financial applications. Publish at top-tier venues while deploying production models serving 200+ enterprise clients at 10B+ API calls per month.

Core Responsibilities

Design and implement novel transformer architectures optimized for long-context document understanding, multi-task learning, and efficient inference
Lead domain adaptation research including pre-training strategies, fine-tuning methodologies, and active learning workflows reducing annotation costs
Publish first-author papers at ACL, EMNLP, NeurIPS, ICML, and ICLR advancing NLP techniques for specialized domains
Develop interpretability methods providing explicit reasoning chains, confidence calibration, and uncertainty quantification for enterprise AI
Build multilingual NLP systems supporting 50+ languages with cross-lingual transfer learning and zero-shot capabilities
Collaborate with engineering teams to deploy research models at production scale with <100ms latency and 99.9% uptime
Mentor PhD-level researchers and ML engineers, establishing research culture balancing innovation with product impact

Required Qualifications

PhD in Computer Science, Computational Linguistics, or related field with 7+ years of NLP research experience in academia or industry
Strong publication record at top-tier NLP/ML conferences (ACL, EMNLP, NAACL, NeurIPS, ICML) with 10+ peer-reviewed papers
Deep expertise in transformer architectures (BERT, GPT, T5, LLaMA) and modern pre-training techniques
Proven track record deploying NLP models to production serving millions of requests with strict latency and accuracy requirements
Expert-level proficiency in PyTorch or TensorFlow with experience training large language models on distributed GPU clusters
Strong understanding of NLP evaluation methodologies, benchmark datasets, and statistical significance testing
Experience with domain adaptation techniques for specialized corpora in legal, medical, financial, or scientific domains

Preferred Qualifications

First-author publications specifically on domain adaptation, few-shot learning, or interpretable NLP
Experience with federated learning, differential privacy, or privacy-preserving NLP techniques
Background in multilingual NLP, low-resource languages, or cross-lingual transfer learning
Contributions to open-source NLP libraries (HuggingFace Transformers, AllenNLP, spaCy)
Postdoctoral research or industry research lab experience (Google AI, Meta FAIR, Microsoft Research)
Understanding of legal, medical, or financial domain knowledge enabling effective model development

Why This Role Matters

Your research will enable lawyers to analyze contracts faster, doctors to extract insights from medical records, and financial analysts to understand regulatory filings. Publish groundbreaking NLP research while deploying models impacting Fortune 500 enterprises and critical industries.

Express Interest

Senior ML Engineer, NLP Infrastructure

Location: Philadelphia, PA or Remote

Compensation: $170,000 - $240,000 + equity

Type: Full-time

Build and scale production NLP infrastructure serving 10B+ API requests per month with <100ms latency. Architect model serving systems, training pipelines, and MLOps workflows enabling rapid iteration from research to production deployment.

Core Responsibilities

Design and implement high-performance model serving infrastructure using TensorRT, ONNX Runtime, and GPU optimization for sub-100ms inference latency
Build distributed training pipelines for large language models spanning multiple GPUs and nodes with efficient data parallelism and model parallelism
Develop MLOps workflows including model versioning, A/B testing, canary deployments, and automated rollback mechanisms
Architect data processing pipelines handling petabyte-scale text corpora for pre-training, fine-tuning, and evaluation
Implement monitoring and observability systems tracking model performance, latency, throughput, and accuracy in production
Optimize model size and inference speed through quantization, pruning, distillation, and efficient attention mechanisms
Collaborate with research scientists to productionize cutting-edge NLP models while maintaining strict SLA requirements

Required Qualifications

BS/MS in Computer Science, Engineering, or related field with 6+ years of ML engineering experience building production NLP systems
Deep expertise in model serving frameworks (TensorFlow Serving, TorchServe, NVIDIA Triton) and GPU optimization techniques
Strong systems programming skills in Python, C++, or Rust with experience optimizing performance-critical code paths
Proven track record deploying NLP models at scale handling millions of requests per day with strict latency budgets
Experience with Kubernetes, Docker, and cloud infrastructure (AWS, GCP, Azure) for ML workload orchestration
Proficiency in distributed training frameworks (DeepSpeed, Megatron-LM, Ray) for large-scale model training
Understanding of transformer architectures, attention mechanisms, and modern NLP model optimization techniques

Preferred Qualifications

Experience with model compression techniques (quantization, pruning, knowledge distillation) reducing inference costs
Background in CUDA programming and custom kernel development for GPU acceleration
Familiarity with federated learning infrastructure and privacy-preserving ML deployment
Contributions to open-source ML infrastructure projects (KubeFlow, MLflow, Ray, Horovod)
Experience building data pipelines for NLP using Apache Spark, Dask, or distributed processing frameworks
Track record optimizing inference latency and throughput for production language models at enterprise scale

Why This Role Matters

Your infrastructure enables real-time NLP applications powering legal research, medical diagnosis support, and financial analysis at Fortune 500 companies. Build systems at the forefront of production ML engineering, solving challenges at the intersection of research innovation and enterprise requirements.

Express Interest

Principal Applied Scientist, Domain Adaptation

Location: Philadelphia, PA or Remote

Compensation: $200,000 - $280,000 + equity

Type: Full-time

Lead domain adaptation research and deployment for legal, medical, and financial NLP applications. Develop novel transfer learning techniques, active learning workflows, and synthetic data generation methods achieving state-of-the-art performance with limited labeled data.

Core Responsibilities

Design and implement domain adaptation strategies for legal contract analysis, medical record extraction, and financial document understanding
Build active learning systems reducing annotation costs by 80% through intelligent sample selection and human-in-the-loop workflows
Develop synthetic data generation pipelines using GPT-4, Claude, and custom generative models for data augmentation
Conduct domain-specific pre-training on corpora including case law, medical literature, SEC filings, and patent databases
Collaborate with legal, medical, and financial domain experts to define evaluation benchmarks and validate model performance
Lead customer engagements for Fortune 500 clients, customizing models for proprietary vocabularies, taxonomies, and use cases
Publish applied research at industry conferences and write technical blog posts demonstrating domain adaptation best practices

Required Qualifications

PhD in Computer Science, Computational Linguistics, or related field with 5+ years of applied NLP experience in specialized domains
Deep expertise in transfer learning, domain adaptation, and few-shot learning techniques for NLP applications
Proven track record deploying domain-specific NLP models for legal, medical, financial, or scientific applications
Strong understanding of active learning, semi-supervised learning, and human-in-the-loop machine learning workflows
Experience with data annotation platforms, labeling guidelines development, and inter-annotator agreement evaluation
Proficiency in PyTorch and HuggingFace Transformers for fine-tuning large language models on custom datasets
Excellent communication skills with ability to engage domain experts, customers, and cross-functional teams

Preferred Qualifications

Publications at ACL, EMNLP, or domain-specific venues (ACL-BioNLP, NLLP Legal NLP, FNP Financial NLP)
Background in legal informatics, medical NLP, or financial text mining with domain knowledge
Experience with synthetic data generation using large language models and controllable text generation
Familiarity with ontology learning, taxonomy construction, and knowledge graph extraction from text
Track record consulting for law firms, healthcare institutions, or financial services companies on NLP projects
Understanding of regulatory requirements (HIPAA, GDPR, legal privilege) impacting NLP deployment

Why This Role Matters

Your work enables NLP breakthroughs in domains where general-purpose models fail. Partner with leading law firms, hospitals, and financial institutions to build custom language models transforming how professionals work with specialized text at scale.

Express Interest