Machine Learning Engineer - Generative Video & Avatar Models

Mindflix AI is a Singapore-headquartered AI company with its India subsidiary and core product team based in Bangalore. We are building an AI-native engagement platform combining agentic AI, generative video, digital avatars, voice, and enterprise automation. Our goal is to help organisations move from generic communication to personalised, contextual, and human-like engagement at scale.

About the Role

Mindflix AI is looking for a Machine Learning Engineer / Research Engineer to work on generative video and avatar model systems.

This role will focus on training, fine-tuning, adapting, and improving models related to talking avatars, lip synchronization, facial animation, identity preservation, and audio-driven video generation.

The candidate does not need prior experience with this exact model category. However, they should have strong hands-on experience training deep learning models, preferably in computer vision, generative AI, diffusion models, image generation, or video generation.

This is a hands-on role for someone who can work with research papers, open-source model code, GPU training pipelines, video/audio datasets, and real-world model quality improvements.

Responsibilities

• Train, fine-tune, and adapt generative video and avatar models.
• Work with video, image, audio, and face datasets for model training and evaluation.
• Build and maintain training pipelines using Python and PyTorch.
• Improve lip-sync quality, facial realism, identity preservation, audio-visual alignment, and temporal consistency.
• Read research papers and adapt open-source GitHub repositories into stable internal workflows.
• Debug model training issues, data pipeline issues, GPU memory issues, and inference-quality problems.
• Run experiments on local GPU machines, cloud GPUs, or GPU clusters.
• Evaluate outputs through visual inspection, internal quality benchmarks, and relevant technical metrics.
• Collaborate with product, engineering, and creative teams to improve avatar output quality for real-world use cases.

What We Are Looking ForMust Have

• 2 to 4+ years of hands-on experience training deep learning models.
• Strong Python and PyTorch skills.
• Experience in computer vision, image/video processing, or generative AI.
• Experience working with generative models such as diffusion models, GANs, transformers, image generation models, or video generation models.
• Familiarity with OpenCV, FFmpeg, or similar video/image processing tools.
• Experience preparing, cleaning, and structuring datasets for model training.
• Ability to read research papers and implement/adapt research code.
• Experience training models on GPUs and debugging GPU-related issues.
• Strong problem-solving ability and comfort working with experimental model pipelines.

Strong Bonus

• Experience with talking-head generation, avatar generation, face animation, lip-sync, or audio-driven video generation.
• Experience with models such as Wav2Lip, SadTalker, MuseTalk, AnimateDiff, Stable Diffusion, diffusion transformers, or similar generative video systems.
• Experience with audio-visual models, speech-driven animation, or facial landmark-based animation.
• Experience improving temporal consistency, identity preservation, expression quality, or video realism.
• Experience with distributed training, multi-GPU training, mixed precision training, or model optimization.
• Familiarity with Hugging Face, CUDA, accelerate, DeepSpeed, or similar model training tools.
• Experience converting research code into repeatable internal pipelines or production-adjacent systems.

Preferred Mindset

• Comfortable working in an early-stage AI product environment.
• Strong ownership mindset and ability to work independently.
• Practical research mindset: able to balance experimentation with usable product outcomes.
• Willingness to work with imperfect data, fast iterations, and evolving technical goals.
• Interest in building avatar and video AI systems that can be used in real-world enterprise products.