AI Engineer

At Prodigal, we are building AI Agents for loan servicing and collections. Founded in 2018 by IITB alumni, our journey began when our founders Shantanu and Sangram faced the antiquated the $13B+ lending/collections industry was mostly using pen and paper and decided to build the first real intelligence layer for debt recovery. Today, we stand at the forefront of a seismic shift in the industry, building Agentic AI applications for loan servicing and collections. Powered by our cutting-edge platform, Prodigal’s Intelligence Engine (PIE), we’re creating the next-generation agentic workforce - one that empowers companies to achieve unprecedented levels of operational excellence and intelligence. With over 100+ enterprise customers across North America and backing from Y Combinator, Accel and Menlo Ventures, we are the fastest growing AI company in consumer finance. About the Role We're looking for an AI Engineer to join the proAgent team and work on the systems that power our autonomous voice agents in live financial conversations. You'll work alongside a small, high-agency team where you will have real ownership over features and components from day one. This is a hands-on engineering role. You'll be deep in the code - building agent logic, integrating speech models, tuning prompts, and debugging gnarly real-time issues. You don't need to have done all of this before, but you need to be the kind of engineer who figures things out fast, takes feedback well, and ships. What You'll Do Agent Loop & Reasoning Build and maintain components of the agentic runtime that powers multi-turn financial conversations - covering payment negotiations, compliance guardrails, and objection handling Implement context management logic that tracks consumer state, conversation history, and business rules across long, branching dialogues Write and iterate on primitives that balance conversational fluidity with structured reasoning - the agent needs to feel human while making verifiable decisions Voice AI Systems Work on our real-time voice pipeline with a target of sub-1s latency across transcription, reasoning, and synthesis Contribute to VAD tuning and turn-taking logic that makes conversations feel natural Help evaluate and integrate speech models (STT, TTS, speech-to-speech) - we currently work with ElevenLabs and Cartesia for TTS, and Deepgram for STT, and are always exploring what's next Debug streaming audio and WebRTC issues in production LLM Infrastructure Write and refine prompts, implement orchestration flows, and contribute to model routing logic Build components of our evaluation framework - helping measure agent quality across conversation quality, empathy, and compliance adherence Stay curious about new models and tools - flag opportunities and contribute to build-vs-integrate discussions What You Bring 2 – 5 years of hands-on engineering experience, with good exposure to building or working with ML or AI systems in production. Solid Python fundamentals - you are comfortable writing clean, maintainable code and debugging production issues. Some experience working with LLMs: prompt engineering, API integrations, or building simple pipelines or agents. High bias for action - you don't wait to be told exactly what to do, and you push yourself to ship rather than over-engineer. Strong fundamentals in Python; familiarity with TypeScript is a plus. Eager to learn in a fast-moving environment, take ownership of your work, and ask good questions. Even Better Exposure to voice AI: speech recognition, synthesis, or telephony systems - even if only through personal projects or coursework Any background or interest in fintech, lending, or collections You've tinkered with Twilio, LiveKit, ElevenLabs, or similar real-time infrastructure You've built a small agentic system - even a side project - that combines LLM reasoning with structured actions Familiarity with streaming protocols, WebRTC, or low-latency system design Why This Role Real ownership f…