AI Researcher

About Harmony Harmony gives every employee an always-on expert for every service request, inside the tools they already use. Deployed inside Slack and Microsoft Teams, its agents draw on a context graph connecting each employee's identity, devices, applications, and history, resolving requests across IT, HR, finance, procurement, and legal automatically, before a ticket ever opens. With 100+ prebuilt agents that deploy in days, not months, and a 70% no-touch resolution rate, Harmony removes the manual work from enterprise service without removing the humans who should own the hard problems. Built for IT and HR teams that are done waiting for their service platform to catch up with the rest of the business. About the role You will join Harmony’s Engineering team to turn large volumes of real-world IT data into insights, models and product features that power our autonomous IT agents. This role sits between data science, applied AI research and engineering. You will dig into tickets, conversations, knowledge articles and system signals at scale, find what matters, and ship it into the product — not leave it in a notebook. You will work closely with engineers, product managers and domain experts. We are building a product, and the best research here is the research customers end up using. What you’ll do Analyze large volumes of unstructured and semi-structured text (tickets, chats, emails, knowledge bases, logs) to uncover patterns, intents, topics and gaps Build semantic analysis pipelines: embeddings, clustering, classification, entity and intent extraction, similarity and retrieval Explore and compare different approaches to complex AI problems, from established NLP techniques to emerging LLM methods, designing experiments to understand what works, why, and under which conditions. Design experiments and evaluation frameworks, including offline metrics, human labeling and LLM-as-judge, and use them to make clear go / no-go calls Write production-quality code and work with engineers to ship research into live product features, then monitor and improve them Partner with product to define the problem, size the opportunity and decide what is worth building Share findings with the wider team through clear write-ups, dashboards and demos Ideal candidate M.Sc. or Ph.D. in Computer Science, Statistics, Mathematics or a related field, or equivalent practical experience 4+ years as a Data Scientist, ML Researcher or Applied Scientist with a proven track record Proven experience analyzing large data volumes, ideally millions of records, with a strong preference for semantic textual analysis (embeddings, topic modeling, clustering, classification, NLP) Research and experiment with LLM-based approaches, including semantic understanding, information extraction, retrieval, and agent evaluation, to solve complex product problems. Willingness and experience taking research into production: you have shipped models or features, not only prototypes, and you enjoy working with engineers to get there Strong coding skills in Python and the data stack around it (e.g. pandas, PySpark, SQL, PyTorch); you write code others can read, test and maintain Product mindset: you start from the user problem, think in impact and trade-offs, and know when “good enough and shipped” beats “perfect and late” Clear communicator who can explain findings to engineers, product managers and leadership Advantage Experience with IT service management or support data (tickets, incidents, knowledge articles) Hands-on work with RAG, vector databases and semantic search at scale Experience with LLM evaluation, agent benchmarking or LLM-as-judge methods Experience with large-scale data processing (Spark, Databricks, BigQuery or similar) and cloud ML platforms MLOps experience: model serving, monitoring, data drift and feedback loops in production Startup experience, or having worked as an early data or AI hire Equal Employment Opportunity Harmony is an equal opportunity employer. We celebrate the diverse backgrounds, perspectives, and experiences that make our team stronger. All qualified applicants will receive consideration for employment without regard to race, color, religion, sex, sexual orientation, gender identity or expression, national origin, age, disability, genetic information, marital status, pregnancy, veteran status, or any other characteristic protected by applicable federal, state, or local law.