SoundsGoodLab
Realtime audio translation
Outcome: stable turn-taking with low jitter under realtime constraints.
Challenge: Keep jitter low and turn-taking stable.
Tradeoff: Latency vs accuracy—tuned chunk sizes to avoid rubber-banding.
Singapore · Open to AI Agentic / Full-Stack Roles
I’m Vidy, a Singapore-based full-stack engineer focused on agentic systems and real-time flows. I ship prototypes with clear guardrails, measurable outcomes, and reliable operation.
Highlights
Reliability first, then latency, then cost. Each build has metrics, tradeoffs, and guardrails.
<2s E2E · <150ms hop latency
WebRTC + translation worker, tuned chunking to avoid jitter.
Comment → Order → RTMP
Event-driven with backpressure, retries, and DLQs.
Guardrailed tool-use
Safe file edits, search/replace, run tasks with explicit gates.
Featured prototypes
Lead with outcomes, then show the engineering decisions.
SoundsGoodLab
Outcome: stable turn-taking with low jitter under realtime constraints.
Challenge: Keep jitter low and turn-taking stable.
Tradeoff: Latency vs accuracy—tuned chunk sizes to avoid rubber-banding.
livey.ai
Outcome: reliable ingestion → order processing → broadcast with retries and replayability.
Challenge: Reliable chat ingestion → ordering → RTMP with external systems.
Tradeoff: Reliability vs speed—backpressure + idempotency over fragile realtime.
Agentic IDE
Outcome: safe tool-use workflows with explicit approval gates and reversible edits.
Challenge: Safe tool-use over editor actions without workspace damage.
Tradeoff: Autonomy vs safety—guardrails kept strict even if slower.
Skills
Experience
Balanced latency vs stability for audio translation; event-driven RTMP flows.
Rapid prototyping with demand validation and feedback loops.
Realtime experiences (HLS/WebRTC), chat systems, payments flows; performance & accessibility standards.
Contact
Fastest via email or LinkedIn. Include context + timelines.
LinkedIn: linkedin.com/in/vidyalfredo
Resume: Open resume