AI/ML Engineer

To be considered, please submit your application through our applicant tracking system: Apply Here.

About the job

We are a small, fast-paced team building AI-first systems for legal-tech, prop-tech, search, and document intelligence workflows.

Our work deals with messy real-world data: scanned files, legal reports, structured and unstructured records, inconsistent names, ambiguous addresses, noisy OCR, incomplete metadata, and complex business rules. The problems are rarely clean, and that is what makes them interesting.

This role sits across the AI/ML lifecycle: curating data, training and fine-tuning models, evaluating performance, deploying systems, and improving them based on real-world feedback.

We are looking for someone curious, rigorous, and hands-on. Prior production experience is not mandatory, but strong fundamentals, learning speed, and clear thinking are essential.

As an AI/ML Engineer, you will:

• Curate, clean, label, normalize, and structure datasets for AI/ML workflows
• Build systems for entity resolution, name matching, address matching, text normalization, and information extraction
• Fine-tune and evaluate models such as BERT-style encoders, seq2seq models, classifiers, embedding models, rerankers, or extraction models
• Design search and ranking pipelines using keyword search, vector search, reranking, heuristics, and learned models
• Work on OCR-backed document understanding pipelines for scanned files, PDFs, and semi-structured documents
• Evaluate performance using metrics such as precision, recall, F1, accuracy, ranking quality, latency, and error distribution
• Analyze false positives, false negatives, edge cases, and model failure patterns
• Deploy models and ML-backed services into production workflows
• Monitor model and pipeline performance over time
• Work with engineering and product teams to convert ambiguous business problems into datasets, experiments, metrics, and working systems

You'll be a good fit if you:

• Have strong programming, data, and machine learning fundamentals
• Are comfortable working with raw, incomplete, and inconsistent data
• Can reason about model performance beyond "the output looks good"
• Can discuss trade-offs using metrics, error analysis, confidence thresholds, and evaluation sets
• Like building systems, not just running notebooks
• Are interested in NLP, search, ranking, entity resolution, OCR, and document intelligence
• Can read documentation, papers, model outputs, and logs to figure things out
• Can move fast without being careless

Relevant experience

• You should have hands-on exposure to some of the following:
• Python for data processing, automation, scraping, ML workflows, or backend services
• Dataset creation, labeling, cleaning, normalization, or annotation workflows
• NLP tasks such as entity extraction, entity resolution, text classification, semantic similarity, or text normalization
• Fine-tuning or evaluating models such as BERT, sentence-transformers, T5, BART, LayoutLM, Donut, or similar architectures
• Embeddings, vector search, reranking, semantic search, Elasticsearch, OpenSearch, Solr, or similar systems
• OCR, document parsing, PDFs, scanned image processing, or visual document understanding
• Model evaluation using precision, recall, F1, ranking metrics, confusion matrices, error analysis, or custom evaluation datasets
• Deploying ML models or ML-backed APIs using FastAPI, Flask, Docker, or similar tools
• LLM API integrations for extraction, classification, reasoning, or data enrichment

You do not need to know everything listed above. We care more about strong fundamentals, learning speed, curiosity, and the ability to build reliable AI/ML systems from messy real-world data.

What we offer

• Work on serious AI/ML problems with direct real-world impact
• Exposure to search, NLP, document intelligence, model evaluation, and production ML workflows
• A high-ownership environment where good ideas are taken seriously
• Mentorship from a product and engineering team that values clear thinking
• The chance to work across the full AI/ML lifecycle, from raw data to deployed systems