Code efficiency matters!
Horizontally scalable media and data infrastructure, engineered in C from the ground up. Two decades in telecom and broadcast, distilled into systems that do more with less — from prototype to production.
scroll
Code efficiency matters!
Horizontally scalable media and data infrastructure, engineered in C from the ground up. Two decades in telecom and broadcast, distilled into systems that do more with less — from prototype to production.
LYLID Labs builds horizontally scalable Internet services that challenge the assumptions of traditional enterprise infrastructure. Where most stacks are assembled from a dozen off-the-shelf services, we design lean, single-binary systems in C — fewer moving parts, predictable performance, and operational simplicity. Two decades of experience in real-time media and high-volume data handling, from DVB headends to internet-scale CDNs.
Headend development, DVB-T/T2 statistics and monitoring, HbbTV and HLS/DASH delivery. Carrier-grade media handling, hardened for 24/7 operation.
Just-in-time HLS/DASH packaging, server-side ad insertion, DRM, live edge nodes and distributed caching — engineered toward 100 Gbit/s per node.
VM and container orchestration, live migration, high availability and disaster recovery — a leaner alternative to heavyweight enterprise platforms.
Sharded, replicated object and data stores with automatic failover, live resharding and cross-region sync. Strong consistency where it matters.
Lightweight push-based agents with crash-safe offline buffering and SNMP/JSON collection — built for fleets, from IT estates to broadcast plants.
Horizontally scalable backends, load balancers, real-time splicers and secure token systems. Designed to scale out, not just scale up.
Everything runs on a shared in-house stack we have been evolving since 2014 and rewrote from the ground up: custom sharding, primary-backup and quorum replication, automatic failover and live resharding, over HTTP/2, HTTP/3/QUIC and mTLS with async event-loop I/O. No nginx, no Kafka, no external coordinator.
Deployable with scp and systemctl restart — minimal dependencies, no container sprawl.
Written, fuzzed and valgrind-audited for systems that run for years, not demos that run for a meeting.
Consistent-hash sharding and replication built in from day one — designed for internet scale, not enterprise PowerPoints.
Performance is measured, not assumed. "It mostly works" was never good enough in broadcast, and it isn't here.