About the job Computer Vision Engineer - Remote from anywhere in Pakistan
Hiring Computer Vision Engineer - Remote from anywhere in Pakistan
Client Introduction:
Our Client Company is revolutionizing social connections by enabling instant video hangouts with like-minded individuals, tackling loneliness one conversation at a time.
Job Description:
The company's promise of a safe, spontaneous one-to-one video Hangout collapses the moment a user encounters explicit content, harassment, or false identity. We need a specialist who can:
- Stop inappropriate visuals as they happen in the video call in the mobile app (300ms end-to-end).
- Verify users in real time (facial verification)detect a face, confirm it matches the account holder, and flag re-registrations by banned users.
- Infer age range and sex (male/female) to enforce age-appropriate matching and improve profile integrity.
- Fuse image, video, and audio cues to enrich the experience (sentiment, engagement, toxicity).
- Do it privately and cost effectively, balancing on device, edge, and cloud inference.
Core Responsibilities
- Design NSFW / explicit content models or integrate best in class moderation APIs.
- Face-in-frame detection and face verification / liveness to ensure the caller is the rightful account owner.
- Watch-list matching to catch banned users attempting to re-register
- Age-range & sex classification from live video to support age gating and gender-based match filters
- Stream video frames (adaptive FPS) from the media pipeline; decide edge vs server inference.
- Trigger autoblur, mute, or disconnect workflows via backend events.
- Apply speechtotext + toxicity/profanity classifiers and combine with visual signals.
- Extract lightweight sentiment, attention, and engagement metrics for future matchquality analytics.
- Build GPU/CPU inference services or on device TFLite / Core ML builds; profile latency, memory, and battery impact.
- Establish precision/recall targets, maintain confusion matrix dashboards, and drive model retraining cycles.
- Minimize frame retention, document data handling flows, and lead bias & fairness checks.
- Integrate detection events with backend & mobile; advise on policy thresholds and appeal flows.
- Collaborate with backend and mobile engineers to integrate detection models into app flow
- Build systems to extract and analyze video frames securely and efficiently
Minimum Qualifications:
- 5+yrs building computer vision or multimodal ML systems in production.
- Mastery of PyTorch, TensorFlow, or equivalent, plus OpenCV/FFmpeg.
- Realtime or near Realtime video processing background.
- Strong skills in model optimization (quantization, pruning, distillation) and performance profiling.
- Understanding of WebRTC/media pipelines and hooking into frame/spoken audio events.
- Proven experience with monitoring, drift detection, and precision/recall reporting.
- Knowledge of privacy regulations and ethical AI guidelines.
Nice to have
- On device ML (Core ML, TFLite, MediaPipe).
- Media server integration (Live Kit, Janus, Stream, Agora).
- Speech/toxicity or voice emotion classification.
- Serverless GPU or ASIC inference (Nvidia Triton, AWS Inferentia).
- Opensource moderation contributions or NSFW/fairness research.
Other Details:
- Location: Remote from anywhere in Pakistan
- Work Mode: Full Time Remote
- Experience: 5 Years
- Benefit: Tax Free Salary
About HR Ways:
HR Ways is an Award winning Technical Recruitment Firm helping software houses and IT Product companies internationally and locally to find IT Talent. HR Ways is engaged by 300+ Employers worldwide ranging from worlds biggest SaaS Companies to most competitive Startups. We have entities in Dubai, Canada, US, UK, Pakistan, India, Saudi Arabia, Portugal, Brazil and other parts of the world. Join our WhatsApp Channel https://shorturl.at/983azto stay updated or visit www.hrways.co to know more.