Services
Ten areas of work, led by voice. Take one on its own, or have them coordinated together under a single engineer.
What I build
Most engagements start with a voice agent and grow into the retrieval, tooling and platform work around it. Architecture and delivery stay with me either way.
Production voice agents on LiveKit, Pipecat or the OpenAI Realtime API. Built for sub-second response, clean barge-in, and behaviour that holds up on a real phone line rather than in a demo video.
Explore AI voice agent development View the AI voice agent case studyLiveKit Agents builds with WebRTC and SIP: self-hosted or LiveKit Cloud, telephony wired in, and per-minute costs a fraction of the managed voice platforms.
Explore LiveKit development View related projectsCustom real-time pipelines in Pipecat, the open-source Python framework, when you need control over every frame between the microphone and the model.
Explore Pipecat development View related projectsSpeech-to-speech agents on the OpenAI Realtime API over WebRTC or WebSockets, with function calling, and an honest read on when speech-to-speech beats a chained pipeline.
Explore Realtime API work View related projectsPut an AI agent on a real phone number. Twilio Media Streams or SIP trunking, inbound and outbound calling, warm transfer to a human, and call recording that satisfies your lawyers.
Explore telephony integration View related projectsYou launched on a managed voice platform, it worked, and now latency, cost or control has become the ceiling. This is the rebuild onto a stack you own.
Explore a platform migration View related projectsRetrieval-augmented generation over your own documents, for chat or for voice. Built so answers come from your data, with retrieval quality measured rather than assumed.
Explore RAG development View related projectsAgents, assistants and generative tooling on OpenAI, Claude and LangGraph, with evaluation and cost control, so behaviour holds up outside the demo.
Explore LLM integration View related projectsThe product your agent lives inside: React and Next.js on Python, Node.js or Golang, with auth, billing, dashboards and deployment owned by one engineer.
Explore platform work View related projectsLearn to build and ship real products with AI, mentored one-to-one. No prior coding background needed.
See how the mentoring works Course detailsCommon situations
If more than one of these sounds familiar, the fix is usually architectural rather than a matter of adding people.
A voice agent with two seconds of dead air before every reply
Callers talking over the agent, and the agent carrying on regardless
Vapi or Retell per-minute costs that stopped making sense at real volume
A voice demo that impressed everyone and never survived real phone lines
An agent that sounds fine until it has to actually look something up
Speech recognition that collapses on accents, names or background noise
A RAG system that retrieves the wrong passages and answers confidently
LLM costs that climbed quietly until someone finally read the invoice
A voice pipeline that works on a laptop and falls over at ten concurrent calls
A contractor handoff that left undocumented code nobody wants to touch
Direct Project Enquiry
Share what you are building and where it is stuck. If a managed platform already solves it, I will say so. I review every enquiry personally and reply within about four hours.
International projects welcome
I work in English by email, video call or through Upwork, from GMT+7 with overlap hours for US and EU teams.
Read the Privacy Policy, Terms of Service and Cookie Policy.