Forward Deployed Engineer

LLR Partners; Philadelphia (Philadelphia County); On-site; USD 150,000 - 230,000; Full time

Job summary

LLR Partners seeks Forward Deployed Engineers to build AI-native products that increase leverage across the firm. You will ship agentic workflows and internal tools, from problem framing to production, collaborating with Investment, Origination, and the Value Creation Team.

This mid-level role offers real ownership: you will design MCP servers, RAG pipelines, and Claude Skills while ensuring governance, observability, and adoption across the organization in a PHILADELPHIA-based office.

Qualifications

• 2-4 years of professional software engineering with at least 1 year shipping production LLM applications to real users.
• Strong Python (async, typing, testing) and JavaScript/TypeScript tooling for web apps.
• Deep hands-on experience with foundation models APIs and SDKs (Anthropic, OpenAI).
• Experience building RAG pipelines end-to-end with embeddings and vector stores.
• Proven track record as a forward-deployed, founding or early engineer in a small team.

Responsibilities

• Ship AI-native internal products and agentic workflows across multiple teams.
• Build the platform layer with MCP servers and RAG pipelines to expose data to agents.
• Ship internal web apps (JS/React/Streamlit) for non-technical users and production-grade tools.
• Contribute to enterprise knowledge graph and semantic search.
• Develop LLM-as-judge evaluations, guardrails, and observable outputs for governance.

Skills

Python

TypeScript

Next.js

Streamlit

LLM APIs

RAG pipelines

Claude Skills

MCP servers

LangGraph

PydanticAI

DSPy

Production LLM apps

Job description

Overview

LLR Partners is hiring two Forward Deployed Engineers to build the AI-native products that generate real operating leverage across the firm - more AUM per head, more decisions per hour, more institutional memory retained in the firm. You will work shoulder-to-shoulder with LLR's teams - starting with Investment, Origination and the Value Creation Team, and expanding across every function - to ship agentic workflows, custom web apps and enterprise knowledge infrastructure from problem framing to production in weeks.

This is a mid-level seat with real ownership. You will not just consume off-the-shelf AI tools; you will build custom MCP servers, reusable Claude Skills, RAG pipelines, and the semantic and judgement layers that make every agent trustworthy at scale.

Accountabilities

Ship AI-native internal products. Build and own the agentic workflows, copilots and internal tools that investment, origination, investor relations, operations, HR, finance and the value creation team every day.

Build the platform layer. Custom MCP servers exposing LLR's data to every agent; RAG pipelines with chunking, embeddings, vector stores, retrieval/generation and evals; reusable Claude Skills that codify LLR patterns.

Ship custom internal web apps. js, React or Streamlit front-ends that put agents in the hands of non-technical users - polished, fast, production-ready.

Contribute to the knowledge graph. Help stand up the enterprise knowledge graph and semantic search that turn LLR's data into one queryable brain.

Build the judgement layer. LLM-as-judge evals, deterministic assertions, guardrails and observability - plus approval flows, confidence thresholds and escalation paths so no agent output reaches an LP, an IC or a portfolio company without a person in the loop.

Bring rigor. Instrument everything - adoption, usage, hours returned - so the value of every agent is measured, not hoped for.

Partner across the firm. Sit with deal teams, origination, IR, operations, HR, finance and the Value Creation Team to identify their highest-leverage workflows and ship for them end-to-end.

Drive AI adoption across the firm. Run regular trainings and office hours, write playbooks, and sit with users until the tool is habitual - an agent nobody uses is a cost, not an asset.

Skills and Requirements

Ability to work in‑person in LLR's Philadelphia office

2-4 years of professional software engineering, with at least 1 year shipping production LLM applications, agents or retrieval systems to real users.

Strong Python (async, typing, testing); TypeScript, Next.js or Streamlit for shipping custom web apps and internal tools.

Deep hands‑on experience with foundation model APIs and SDKs (Anthropic, OpenAI) - tool use, function calling, structured outputs and prompt engineering.

Built RAG pipelines end‑to‑end - chunking, embeddings, vector stores (pgvector, Pinecone or similar) and retrieval and generation evaluation.

Built custom MCP servers and reusable Claude Skills - not just consumed them. You understand the protocols, can design new integrations, and know when to reach for a Skill vs. an MCP server.

AI‑native engineer. Daily fluency across Claude and ChatGPT ecosystems - Connectors, Claude Code, Codex, Cowork - and agentic frameworks (LangGraph, PydanticAI, DSPy) shipped in production. You know the tradeoffs and pick the right tool per problem.

Track record as a forward‑deployed, founding, or early engineer on a small, high‑ownership team - you've worked directly with non‑technical users on real problems.

Experience designing for regulated environments - data classification, PII handling, scoped access, information barriers and audit logs on every agent action.

Nice to Have

Experience inside private equity, financial services, consulting or another regulated, document‑heavy environment.

Comfort in an Azure environment - including familiarity with Azure AI Foundry - with modern deployment platforms (Render, Vercel, Supabase) and Git‑based workflows.

Experience building harnesses for agents - either using a harness framework or standing up your own.

Knowledge graph or GraphRAG experience - bonus for enterprise search architectures at scale.

Experience communicating technical work to non‑technical stakeholders through writing, decks and live demos.

Flexibility with project management and workflow tools (JIRA, Trello, Linear, Asana or similar).

Working knowledge of PE‑stack data (PitchBook, SourceScrub, Grata, Allvue, Chronograph) and the deal lifecycle - IC memos, LP reporting, fund structures - enough to build useful tools without a translator.

Curiosity about and informed perspective on the evolving AI and agent ecosystem.

Awareness of token economics and inference cost - model selection, prompt caching, routing small vs. frontier models by task.

Experience extracting structure from messy documents - PDF parsing, table extraction, meeting transcripts, email threads.

Observability tooling for agents in