Superduper is looking for a Machine Learning Engineer to architect the real-time platform that transforms noisy social feeds into clear token insights. You will power a fast, intuitive product by ingesting data from every channel, building advanced NLP models, and delivering sub-second API responses.
What You'll Do
- Build and optimize low-latency, high-throughput APIs that expose real-time token mindshare and sentiment metrics to downstream clients.
- Design and implement real-time sentiment-analysis and NLP pipelines for social feeds, covering ingestion, tokenization, entity extraction, and sentiment scoring.
- Develop and train ML models, starting with pre-built services and advancing to custom transformer architectures, to continuously improve the accuracy and relevance of sentiment signals.
- Collaborate with cross-functional teams to roadmap and deliver new ML-driven insights and to design intuitive consumer-facing dashboards and alert systems.
- Ensure data security and compliance, particularly around user-generated content, API keys, and any PII in social media streams.
- Maintain code and model quality: author clean, efficient, and maintainable code; implement comprehensive testing and debugging; and lead code reviews, share best practices, and mentor teammates.
What We're Looking For
- 5+ years in ML or data engineering roles, building production-grade NLP or sentiment systems.
- Proven track record building low-latency, high-throughput data pipelines and APIs using Go, Python, or similar.
- Hands-on NLP experience with both pre-built services and custom transformer models.
- Strong grounding in evaluating NLP models using classification and ranking metrics, and experience running A/B or offline benchmarks.
- Proficient with MLOps and training infrastructure, including CI/CD, hyperparameter tuning, and model versioning.
- Strong social media data extraction and scraping skills at scale.
- Experience with real-time streaming systems and ingesting high-velocity data.
- Deep data-engineering expertise across Postgres, Redis, InfluxDB, and ClickHouse.
- Experience deploying microservices in production using Docker and Kubernetes.
- Skilled in setting up observability and alerting pipelines, including model drift detection.
- Experience with real-time ML inference and model serving frameworks for low-latency applications.
- Experience designing feedback loops, active learning, or user-in-the-loop systems to continuously improve model relevance.
- Experience with Git-based workflows and integrating model training into CI/CD pipeline.
Nice to Have
- Experience fine-tuning large-scale transformer models and prompt-engineering for sentiment tasks.
- Background building active-learning and annotation pipelines to bootstrap training data.
- Familiarity with semantic search or vector databases for topic modeling and similarity queries.
- Familiarity with crypto markets, order books, and risk-management frameworks.
- Familiarity with anomaly-detection methods for streaming text and time-series data.
- Experience developing EVM smart contracts with Solidity and modern toolchains.
- Experience with real-time subscription frameworks or gRPC streaming for live data updates.
Technical Stack
- Languages: Go, Python, Solidity
- ML & NLP: AWS Comprehend, Hugging Face, PyTorch, TensorFlow, BERT, GPT
- MLOps & Pipelines: MLflow, Kubeflow, Airflow, Scrapy, Playwright, Kafka, RabbitMQ
- Infrastructure & Data: Postgres, Redis, InfluxDB, ClickHouse, Docker, Kubernetes, Elasticsearch, FAISS, Pinecone
- Serving & Observability: TorchServe, Triton, BentoML, Prometheus, Grafana
- Tools & Protocols: Git, Foundry, Hardhat, GraphQL, WebSockets, gRPC
Team & Environment
You will collaborate with cross-functional teams including frontend, design, marketing, and product. We are a high-performance team in a fast-paced, dynamic, and innovation-focused environment. We embrace a scrappy start-up culture, see ambiguity as opportunity, and treat speed and agility as key competitive advantages.
