AI chatbot for Meta & Instagram
Live conversational agent on Meta's surfaces — the same real-time intent handling and handoff logic a voice agent needs, minus the audio codec. 110+ hours and running.
- 110+hours to date
- Activein production
← All portfolios
05Portfolio · Voice AI
The same production discipline, pointed at audio
Voice systems are streaming systems: latency budgets, buffering, backpressure and moderation in the request path. My delivered base is the conversational and agent work below — the pipeline discipline transfers directly.
- 3,000+hours across AI contracts
- 60+contracts delivered
- 5.0★typical rating
01Delivered work
Projects on the record
Client names withheld under contract — every figure is as recorded on the Upwork profile.
Executive assistant on self-hosted vLLM
Latency-sensitive model serving inside a client's own infrastructure — the serving and streaming discipline that voice pipelines are built on.
- Gemma + vLLMstack
- Activein progress
Straight answer: my shipped voice-specific contracts are thinner than my RAG and agent record. What transfers is the streaming and serving discipline above — if your build is voice-first, I'll tell you in the first call whether I'm the right fit for it.
02Systems of this class
Recognise the shape
Well-known products of the same class — unaffiliated reference points, so a first conversation starts from a shape you already know.
Need Voice AI work like this?
Send the problem and where it's stuck. I'll tell you honestly whether I'm the right person and what it would take — usually within one business day.