Software Engineer, AI Product Engineering

73 Strings•New York, NY

About The Position

73 Strings is an innovative platform providing comprehensive data extraction, monitoring, and valuation solutions for the private capital industry. The company's AI-powered platform streamlines middle-office processes for alternative investments, enabling seamless data structuring and standardization, monitoring, and fair value estimation at the click of a button. The company is seeking software engineers at mid-level and senior to build the AI features clients use to run post-investment operations, such as extracting financial data from documents, monitoring portfolios, and producing auditable valuations. AI is central to the company's operations and product offerings, influencing how systems are built and what is shipped. This involves writing deterministic machinery around probabilistic components, determining appropriate uses for agents, and building evaluation and guardrails to ensure finance professionals can trust the output.

Requirements

Strong software engineering experience with production ownership.
Full-stack capability with backend depth: ability to design services, reason about failure modes, and build the user interface.
Demonstrated system design ability: ability to take ambiguous problems, propose architectures, and explain trade-offs.
Strong practical understanding of domain-driven design and the judgment to use lighter approaches when appropriate.
Language-agnostic engineering strength; productive in at least one of TypeScript, Java, or Python, and willing to work across all three.
Hands-on experience building with LLMs or other probabilistic components in a product context (prompting, context design, retrieval, structured output, tool/function calling, agent orchestration, evaluation).
A clear point of view on where AI belongs in a system, formed by production experience.
Comfort with modern delivery practices: automated testing, CI/CD, containers and Kubernetes, and observability.
Strong business sense: understanding of users and business impact, and ability to make trade-offs accordingly.
Self-direction: ability to find important work, scope it, and drive it independently.
Questioning mindset: ability to identify and challenge assumptions, unnecessary complexity, or poorly defined metrics.
Belief that a feature is not done until its output is trustworthy without exhaustive checking.
Comfort holding both the domain model and current sprint delivery realities simultaneously.
Enjoyment of working with product leadership rather than waiting for a finished spec.
Care about accurate numbers, understanding their importance in the financial industry.
Understanding the difference between pragmatic and sloppy, and commitment to the former.

Nice To Haves

Experience in a fast-growing company where roadmap, team, and requirements change frequently.

Responsibilities

Design and ship user-facing features built on LLMs, retrieval, extraction models, and agentic workflows, from prototype to client-dependent features.
Own the full feature path: data input, model/agent processing, deterministic validation, and a human-reviewable interface.
Build evaluation harnesses, regression suites, and offline test sets to track feature improvements.
Design for failure modes of probabilistic systems, including hallucinated values, silent low-confidence outputs, prompt injection, and drift after model upgrades.
Instrument features for production quality (confidence, correction rates, human overrides, latency, cost per run) and act on the data.
Decide which system parts must be deterministic (e.g., arithmetic, audit trails, permissions, reconciliation) and which can be probabilistic.
Keep probabilistic components within deterministic boundaries using schema-constrained outputs, validation, and reconciliation against source data.
Push back on agentic designs that add nondeterminism without value, providing reasoned justification.
Make reasoning visible in design documents for future engineer understanding.
Design services and system boundaries reflecting the business domain.
Apply domain-driven design principles: bounded contexts, aggregates, ubiquitous language, and context mapping.
Model the private capital domain carefully (funds, portfolio companies, positions, valuations, etc.) and maintain model integrity.
Own services end-to-end: implementation, tests, CI/CD, deployment, observability, and on-call responsibilities.
Work directly with product leadership to shape product development, challenge assumptions, and scope features.
Partner with Data Engineering, ML, and other product verticals on contracts and interfaces to avoid redundant problem-solving.
Drive technical initiatives across the domain, such as shared patterns for AI features, evaluation tooling, domain models, or service consistency standards.
Raise the bar through design review, code review, and mentorship.
Write clearly and produce design documents that survive review.
Use AI as a working tool for exploring code/data, generating test cases, drafting designs, triaging issues, and writing documentation.
Build or configure tools (agentic workflows, evaluation harnesses, custom tooling, MCP integrations, automated review/triage) to improve engineering speed or reliability.
Verify and test AI-generated code, understanding how it was verified.
Apply the same scrutiny to AI-generated code as to junior engineer code.
Share effective AI techniques with the team.