Companies in this catalog do not charge applicants. Never send money to apply.
Principal Data Scientist, AI (Architect-Level Scope)
About KLDiscovery
KLDiscovery is a global eDiscovery and legal technology provider serving large law firms, corporate legal departments, and government agencies. We build and operate the products and services that legal teams rely on to manage, process, and review case data at scale. With operations across multiple countries and a client base that includes AmLaw 200 firms, we handle some of the largest and most complex matters in the industry.
About the Role
We're hiring our most senior AI practitioner — someone with genuine data science rigor who also wants to build. In this role, you'll set the scientific and architectural direction for gen AI and ML at KLDiscovery : designing the evaluation methodology that proves our AI holds up under legal scrutiny, choosing and validating the models and retrieval systems that power our products, and building the hardest parts of that system yourself.
This is one of the most interesting data science problem sets in enterprise software. You'll work with terabytes of real-world legal data — emails, contracts, chat transcripts, images, video, depositions, and regulatory filings — from some of the largest litigation and investigation matters in the world. The work is hard in ways that matter: documents are messy and adversarial, the stakes are real (privilege, defensibility, attorney work product), and getting the evaluation wrong has real consequences. AI that can surface key people, themes, and timelines in hours instead of weeks, or pre-classify millions of documents for relevance and privilege with results that survive scrutiny, directly changes the economics of how legal matters get resolved. We build solutions that turn data into evidence the legal system can trust.
This is a builder-first role, not a research-only or advisory one. You'll design experiments and evaluation frameworks that prove our AI is defensible in court, build the retrieval and agent systems that reason over case data, and personally ship the hardest, most novel parts of that system. You'll set strategic direction across Nebula, our eDiscovery platform, and CS & Operations, then prove out the science and the architecture by building it yourself. Not a role for anyone stepping back from the keyboard, and not a role for anyone who treats evaluation as an afterthought.
We offer competitive total compensation that includes base pay, bonus potential, equity, inclusive benefits, wellness programs, and perks. We use market and industry data to inform pay decisions while considering geography and labor markets, individual experience, and business needs. Individual compensation will vary, although a reasonable estimate of the current annualized base pay range for this position is $190,000 to $240,000.
Job location: Remote (candidate must be based in the United States)
Key Responsibilities
Own the science and the architecture, and build both personally. Design the experimentation and evaluation methodology that determines whether our AI's outputs are accurate, consistent, and defensible — then define the end-to-end gen AI architecture across Nebula and CS & Operations to deliver on it: LLMs, agent harnesses, RAG, vector search, embeddings, and model selection and triage. Build the hardest parts personally: prototype agent loops, design and run the evaluations that validate them, tune retrieval, and ship the shared infrastructure that powers AI Case Explorer (case overviews, timelines, key people and themes, PII surfacing, and Agent chat), AI Agent Review (pre-classifying relevance, privilege, and key issues, shipping MLP), and CS & Ops tech-enablement as part of our central work orchestration system.
Own evaluation rigor, AI/MLOps, and telemetry end-to-end. Design the statistical and experimental methodology behind our evaluation harnesses — the standard our outputs have to meet to hold up under legal scrutiny — and own the infrastructure that enforces it: model deployment and versioning, eval pipelines, drift and quality monitoring, cost and latency telemetry, and prompt and agent observability. Define and implement how we select, triage, and route across models (Azure OpenAI, Anthropic, open-source, fine-tuned), manage vector databases and retrieval, and evolve our evaluation methodology and agent harness as the frontier moves.
Lead the practice from the front. Set the technical and scientific bar by building, not by reviewing. Partner with Engineering, Product, and Data Science leadership to translate that rigor into shipped product. Raise the bar on both AI engineering and evaluation discipline, mentor senior ICs through hands-on technical leadership, build and maintain relationships with model and infrastructure vendors, and represent KLD's AI strategy directly with customers, partners, and at industry events.
What You Bring (Required)
7+ years in data science, applied AI/ML, or ML engineering, with recent hands-on experience as a senior or principal-level practitioner in the gen AI era
Real training in statistics, experimentation, or applied research you know how to design an evaluation that actually tests what you think it tests, not just one that looks reasonable
Proven track record architecting and personally building enterprise gen AI systems in production with measurable customer impact Experience with consumer-facing or B2B customer-facing AI products — systems that real external users or customers depend on, not only internal tooling Builder at heart: still writes code, runs experiments, ships, and tunes prompts and evals, and wants to keep doing so as a leader Deep expertise across the modern gen AI stack: LLMs, agents, RAG, vector databases, embeddings, search, and evaluation harnesses Hands-on experience designing system-of-systems AI pipelines spanning search, retrieval, agent harnesses, and model selection/triage Strong proficiency with the Microsoft AI stack: Azure OpenAI, Azure AI Foundry, Azure AI Search, and supporting Azure infrastructure Excellent technical leadership skills; demonstrated ability to influence architecture and methodology decisions across product, engineering, and data science Strong communication skills, including explaining evaluation results and architecture trade-offs to executive and customer audiences Career experience spanning both a larger, established technology company and a smaller company or startup — comfortable operating with both institutional rigor and startup pace Mentor team members and contribute to a culture of continuous improvement Nice to Have (Preferred) Advanced degree (MS or PhD) in Statistics, Machine Learning, Computer Science, or a related quantitative field Peer-reviewed publications, technical writing, conference talks, or other evidence of contributing to the field, not just building within it Background building agentic systems with tool use, planning, and multi-step reasoning in production Prior experience setting up AI governance and evaluation harnesses in a regulated or high-stakes domain Open-source contributions or other evidence of being a recognized builder in the AI community A demonstrated track record of sticking with hard, ambiguous problems over long timeframes rather than pivoting away when the first approach doesn't work.
