Applied Scientist

About the Company

A stealth-mode AI company affiliated with B Capital is developing decision intelligence solutions tailored for private capital markets. Operating discreetly with locations in Dublin, Singapore, and New York, the company focuses on delivering platforms that transform how private equity and venture firms across Europe, the UK, and the US learn from their investment decisions.

Role Overview

This position demands leading applied research to convert convoluted, real-world business data into systems capable of reliable extraction, retrieval, and reasoning — supporting high-value investment decisions. The role focuses less on novel AI architectures and more on overcoming challenges related to messy data sources, permissioned retrieval, long-term evaluation, and integrating model performance closely with real-world outcomes.

Key Responsibilities

• Develop pipelines to extract and ground terms, KPIs, financials, covenants, and commitments from a variety of documents including investment committee memos, LP letters, Excel models, presentations, PDFs, and emails — ensuring all claims are traceable to original documents, versions, authors, and access permissions.
• Own and optimize retrieval quality across heterogeneous and permissioned historical corpora, ensuring precise document selection relevant to the decision-making process.
• Create systems to infer structured decision records, capturing reasoning, assumptions, and confidence from existing workflows to enable effortless data capture.
• Design and implement evaluation infrastructure including defining correctness criteria for extraction and decision derivation, developing gold-standard datasets with domain experts, and establishing routine regression testing.
• Work on calibration and outcome scoring by linking recorded decisions to realized outcomes, assessing calibration at personal and organizational levels, and identifying systematic divergences between confidence and results.
• Maintain model-agnostic development to enhance the platform's value with advances in AI models without dependency on specific vendors.
• Engage closely with end users by observing investment committee and diligence activities to incorporate real-time insights into system design.

Candidate Profile

Essential Qualifications

• At least 8 years experience building production-grade applied ML/AI systems with recent involvement in large language model-based solutions involving retrieval, agents, structured extraction, fine-tuning, or evaluation.
• Expertise in large language models, neural networks, and retrieval systems including their industrial application, failure modes, and evaluation methodologies.
• Advanced proficiency in Python along with contemporary applied AI technology stacks.
• Practical and extensive usage of LLM-assisted coding tools as integral to development workflows, with informed perspectives on their advantages and limitations.
• Proven experience converting complex, unstructured data from real-world environments into dependable systems, handling data cleaning, entity resolution, and thorough error analysis.
• Strong commitment to meticulous evaluation processes, including building evaluation infrastructure capable of identifying issues missed by informal testing.
• Ability to thrive in ambiguous environments with a strong drive to deliver results in an early-stage, technically formative setting.
• Exceptional written communication skills for diverse audiences including engineers and investors.

Highly Desirable Attributes

• Background in financial services, investment or legal technology areas featuring critical documents, permission constraints, and low tolerance for errors and hallucinations.
• Experience with knowledge graphs, entity resolution, or temporal and bitemporal data modeling.
• Academic research credentials such as published papers, significant open-source projects, or a PhD in machine learning, natural language processing, statistics, or related fields. Equivalent hands-on expertise is also acceptable.
• Familiarity with multi-chain processing, agent tooling, or developing AI interfaces for third-party integrations.
• Expertise in calibration, forecasting, causal inference, or decision science.

Not Required but Valued

• Prior experience in private equity is not necessary; enthusiasm for understanding investment decision-making processes is paramount.

Additional Information

This role offers the unique opportunity to tackle a protected research challenge where outcome-labeled institutional judgments only accumulate after capture begins, safeguarding the dataset's exclusivity. The position includes founding-level responsibilities, influencing technical direction within a pioneering environment focused on making a firm's contextual data fully AI-accessible, governed, and traceable.

The application process is multi-staged with initial screening, technical deep dives, live practical exercises, system design evaluations, and discussions with founders, aiming for closure within three weeks.

The employer is an equal opportunity organization committed to inclusivity and provides reasonable accommodations upon request during hiring.

Confidentiality measures require further product and company details to be shared under NDA at advanced stages.