ML Engineer (Onsite, Lahore, Remittance Salary)

Requirements:

• 2-3 years hands-on with CV or generative models (PyTorch preferred).
• Proven ability to read/understand GitHub repos, tweak code, and retrain or finetune models.
• Comfortable serving models via Python APIs (FastAPI/Flask) and using Docker for reproducible runs.
• Strong debugging mindset; you can profile hotspots and trim latency pragmatically.
• Git, Linux, and basic cloud/GPU know-how.
• Model optimization: quantization (INT8), ONNX Runtime / TensorRT, pruning, batching tricks.
• CI/CD for ML (GitHub Actions/Jenkins), simple model versioning.
• Kubernetes or serverless deploys; monitoring stacks (Prometheus/Grafana/ELK).
• Audio/TTS or phoneme alignment experience.
• Real-time streaming protocols (WebSockets, RTMP/WebRTC) or VTuber/graphics pipeline exposure.

Responsibilities:

• Own the end-to-end model pipeline: integrate/replace open-source or research models (e.g., MuseTalk variants) as they emerge.
• Build a low-latency inference service (FastAPI or similar) that handles live audio/video streams.
• Implement basic preprocessing (OpenCV/PIL/MediaPipe) for frames, audio alignment, and gesture inputs.
• Track runtime issues, fix bugs, and iterate quickly on output quality (lip timing, jitter, desync).
• Collaborate with backend/streaming engineers to plug models into the Twitch pipeline.