colombianinsider.com · 2 hours ago
Co-Founder / Lead AI Engineer (Real-Time Audio Processing for Speech)
United States
Full-time
Remote
Lead/StaffColombian Insider is building a commercial-grade, fully offline, real-time voice conversion communication system for Windows. They are seeking a Co-Founder / Lead AI Engineer to own the entire AI lifecycle, from data curation and training to optimizing the C++ inference engine, with a commitment to delivering the product within 6-12 months.
Staffing & Recruiting
Responsibilities
Own the entire AI lifecycle from data curation and training to optimizing the C++ inference engine
Implement the C++ inference wrapper using the ONNX Runtime C++ API
Clean and curate speech datasets and run automated QA metrics
Optimize models for inference speed on older GPUs
Qualification
Required
Core AI/ML: Deep experience with PyTorch and Speech Synthesis architectures. You must have practical experience with RVC (Retrieval-based Voice Conversion), Soft-VC, HiFi-GAN, and Content Encoders like ContentVec or HuBERT
Inference Optimization: Proven experience converting PyTorch models to ONNX. You must understand Static INT8 Quantization and how to optimize models for inference speed on older GPUs (avoiding dynamic shapes, managing receptive fields)
C++ Implementation: Proficiency in Modern C++ (C++17/20). You will not just train models in Python; you must implement the C++ inference wrapper using the ONNX Runtime C++ API
Data Science: Experience cleaning and curating speech datasets (e.g., L2-ARCTIC, LJSpeech) and running automated QA metrics (WER, MOS/NISQA)
Environment: Must have access to a Windows environment (native or dual-boot) to test DirectML compatibility
Financial Runway: You have the financial stability to work without a paycheck until we reach revenue
Reliability & Grit: You are a finisher. You do not flake when technical challenges arise. You have a track record of seeing projects through to completion
Communication: You are responsive and communicative. As a remote partner, 'going dark' for days is not an option
Extreme Ownership: You don't wait for tickets. You understand the high-level goal (Low Latency + Naturalness) and you proactively solve problems
Partner Mentality: You are not looking for a boss; you are looking for a business partner. You care about the product's success as much as the code quality
Preferred
DirectML Experience: Specific experience optimizing ONNX models for Windows DirectML to support Intel/AMD/NVIDIA GPUs simultaneously
DSP Knowledge: Understanding of Digital Signal Processing (Circular buffers, FFTs, overlap-add, crossfading) to better collaborate with the Audio Systems Engineer
Audio Quality Assessment: Familiarity with objective audio metrics (PESQ, STOI, NISQA) and how to automate them
Version Control: Disciplined use of Git (Branching strategies, PRs)
Benefits
Equity (Ownership)
Revenue Share
Company
colombianinsider.com
2-10 employees
Funding
Current Stage
Early StageCompany data provided by crunchbase