Flam
Software EngineerBackend · AI · Streaming
End-to-end engineering across an AI-driven talking-avatar platform, adaptive media streaming, AR/CV tooling, and the access-control + payments layers underneath.
i.AI Avatar Platform
- Architected a talking-avatar product enabling high-fidelity face and voice cloning.
- Integrated RAG for context-aware responses, cutting inference latency and lifting conversation quality.
- Engineered GPU workflows for fine-tuning and end-to-end avatar creation journeys.
ii.Media & Streaming
- Built an adaptive-bitrate pipeline (HLS/DASH) on FFmpeg + Shaka Packager with auto manifest generation.
- Introduced a NATS JetStream workflow — fault-tolerant, horizontally scalable media processing.
iii.Computer Vision & AR
- AR image-trackability rating for ARCore + ARKit, powering reliable campaigns.
- Vector-search image retrieval on local descriptors — 1M+ corpus served in <50 ms.
- Image-similarity + duplicate detection to clean up the asset pipeline.
iv.Platform & Access Control
- Microservices and clean CRUD APIs for modular scaling and consistent service contracts.
- RBAC via OpenFGA, supporting deep permission hierarchies across services.
- Multi-tier payment + credit system mapping licenses and entitlements onto campaigns.