Applied AI Engineer

Sapien

WHO WE ARE Sapien is rethinking how finance teams operate in the age of AI. We’re building toward an autonomous CFO—systems that run company financials end-to-end. Our platform analyzes complex financial data in real time to increase decision cadence, prevent costly mistakes, and surface value. Sapien has caught multi-million-dollar errors, saved thousands of jobs, and returned significant dollars to customers’ bottom lines. We partner with traditional businesses—manufacturing, healthcare, restaurants, and large enterprises—to deeply understand and automate their financial workflows. We’re HQ’d by Madison Square Park in NYC and backed by General Catalyst, Neo, and top operators from Google, OpenAI, Microsoft, Ramp, and Stripe (over $9M raised). THE ROLE You'll build the AI agent capabilities that power Sapien's autonomous finance operations. This means designing novel architectures for reasoning over complex financial data, implementing verifiable and observable agent workflows, and building systems that learn and adapt to each company's unique operations. This is a research-meets-product role. You'll work on cutting-edge agent capabilities—from observability and library learning to semantic search and multi-modal parsing—and ship them directly into production for customers. WHAT YOU'LL DO * Design and implement agent architectures that enable observability, human-in-the-loop verification, and precise context control across complex financial workflows. * Build library learning systems that reduce LLM dependencies by learning reusable patterns for planning, code generation, and data localization from customer interactions. * Create graph-based company representations and develop efficient search methods using embeddings, semantic clustering, and custom retrieval strategies. * Build multi-modal parsers that unify diverse financial data sources (Excel, ERPs, CRMs) into coherent, queryable schemas that agents can reason over. * Design benchmarking and evaluation suites that quantify Sapien's accuracy, reliability, and business impact across different customer workflows. WHAT WE'RE LOOKING FOR * Strong algorithmic thinking. Demonstrated through ML research, competitive programming, mathematics, or building novel systems from scratch. * Experience with modern agent frameworks, LLMs, and AI systems: fine-tuning, retrieval augmentation, tool use, or agentic architectures. * Comfort working end-to-end: from implementing research ideas and prototyping architectures to deploying production systems and iterating on real customer feedback. HOW YOU WORK (VALUES WE CARE ABOUT) * Adaptability: You thrive on diverse, open-ended problems. One day you're implementing a new agent architecture, the next you're debugging a customer data issue—and you bring the same intensity to both. * Ownership: You take problems from research paper to production. You don't wait for direction—you identify what needs to happen and drive it to completion. * Opinionated: You have strong technical perspectives on what works and what doesn't. You push back on approaches you disagree with and make the team's thinking sharper. * Mission-driven: You're energized by building AI that transforms how businesses run. You connect technical decisions to real customer outcomes and care deeply about the impact. * Collaborative excellence: You share knowledge, give thoughtful feedback, and invest in making the team better. You bring new ideas from papers and discussions to elevate everyone's thinking.

Seamflow

ABOUT THE JOB We’re removing one of the last great barriers to progress in the physical world. While software moves fast, industries like aerospace, healthcare, and energy are still slowed by outdated, manual systems. We are on track to raise our Series A, backed by a 14× oversubscribed $4.5m seed round from Tier-1 VCs such as Coinbase, Reddit, Klarna, and Spotify. Our world-class team includes former unicorn founding engineers and talent from X, Google, Amazon, and Yandex. We value people who want to win and thrive in high-ownership environments. We’re building the system that lets complex, real-world work move at software speed. WHO WE ARE LOOKING FOR We're looking for an Applied AI Engineer who thinks in user workflows and model behavior, not benchmarks or isolated prompts. This is a role for someone who wants to decide what AI should do, build it end-to-end, and put it in users' hands. You will fit right in if you have experience with: * Applied AI — Design and build LLM-powered features that solve real workflows, from prompt to production * Model Behavior — Reason about why models fail, when to use which model, and how to make them reliable in production * Evals & Iteration — Build evals that actually catch regressions and drive improvements, not vanity metrics * Agents & Tooling — Design multi-step agents, tool use, and orchestration for complex domain workflows * Document Intelligence — Extract structured data from messy real-world documents (PDFs, scans, forms, tables) * Product Judgment — Be opinionated about what AI should do, what it shouldn't, and when "good enough" ships WHAT YOU WILL DO * Ship from Day 1 — Deploy AI features to production immediately and own a real user-facing surface area * Own it end-to-end — From problem definition to prompts, evals, monitoring, and iteration * Solve real problems — Sit in customer calls, look at real documents, debug bad outputs, and make the system better WHAT WE ARE LOOKING FOR TECHNICALLY * Strong experience with Python and modern LLM tooling (OpenAI/Anthropic/Azure OpenAI SDKs, structured outputs, function calling) * Experience building production AI systems beyond a wrapper around a chat API * Comfortable with evals, prompt engineering, and reasoning about model failure modes * Ability to work across the stack to integrate AI into real product workflows (TypeScript/React frontend, Python backend) * Ability to reason about complex domain workflows, not just prompt-response Strong plus if you have: * Founder or ex-founder experience * Early-stage or stealth startup background * Founder fellowships or similar environments (EF, YC, Antler, etc.) * Experience building AI-powered products in production * Experience working where requirements were unclear and evolving YOUR FIRST MONTH * Ship an AI feature to production from day one * Own problems end-to-end, from prompt to eval to deployment * Do what needs doing, including looking at hundreds of real outputs, debugging edge cases, and customer interaction WE OFFER * Pace and purpose — In-person in London (visa sponsorship available) * No friction — Relocation, visas, meals, and tools fully covered * A tight-knit team — High standards, high trust, no politics

MTR Legal

WIR sind MTR Legal Rechtsanwälte. Mittelständisch und international. Wir beraten im gesamten Familienrecht. Ein starker Wegbegleiter für nationale und internationale Mandanten. Wir verfolgen nicht nur Leistung, sondern schaffen auch Raum für Persönlichkeiten, um gemeinsam an einem Strang zu ziehen. Denn wir wissen: Mit Ihnen bringen wir unser Beratungsniveau für unsere Mandanten auf das nächste Level! DU Was du dafür können musst * LLM/RAG in Produktion (nicht nur Demo) * Dokumentenfokus (OCR/Extraktion/Klassifikation) oder sehr nah dran * Engineering: Versionierung, Repro, Tests, Monitoring * Evaluation-Mindset: Qualität messen & Regression verhindern Was am Ende stehen muss (Deliverables) * DocTypes + Extraktionsschemas (Top 10) inkl. Validierungen * OCR/Extraktionspipeline (Edge Cases: schlechte Scans, Tabellen, Stempel) * RAG/LLM: Summary + Mandanten-E-Mail-Draft + Action Items * Confidence/Checks: Thresholds, Rule/LLM-Checks, Human-in-the-loop Gates * QA/Eval: Golden Set, Metriken (Precision/Recall), Prompt/Model Regression Tests CONNECTING THE DOTS Tech (Orientierung) OCR/Extraction, LLM/RAG, Embeddings, Prompting/Templates, Eval Harness (Golden Sets), Monitoring/Drift. Integration in Workflow/Queue. Rahmenbedingungen * 100% Remote (CET/Europa ideal) * Projektbasis, Start asap, Auslastung nach Vereinbarung PERSPECTIVE Produktionsreife Document Intelligence: DocType → Felder → Summary → Draft → Checks. Wichtig: messbare Qualität, Regression-Schutz, sichere Gates (Human-in-the-loop).