Your WebRTC calls should just work.
Let’s find out why they don’t.

I help engineering teams fix connection failures, scale real-time media, and cut TURN and CPaaS costs — architecture reviews, debugging sprints, and ongoing advisory from a specialist who’s shipped it before.

20+ years building real-time systems · usually scheduled within 48 hours

sound familiar?

The failure you’re seeing has a name — and a fix.

  • “Calls fail behind corporate firewalls.”

    Symmetric NAT and locked-down networks are eating your sessions. The fix is a correctly deployed TURN layer and an ICE configuration that actually completes — I design and verify both.

  • “Audio degrades as we scale.”

    Jitter, packet loss, and starved jitter buffers show up long before your dashboards do. I run call-quality forensics on real traces and fix the pipeline, not the symptom.

  • “Our TURN and CPaaS bills are exploding.”

    Relay traffic you don’t need, vendor minutes you shouldn’t be buying. An architecture pass usually finds 30–60% of the spend hiding in relay ratios and topology choices.

  • “The mesh dies past six peers.”

    Full mesh has a hard ceiling. The answer is the right SFU — Janus, mediasoup, LiveKit, or Pion — sized for your traffic, with simulcast where it pays. I’ll pick it with you and prove it out.

what I do

Deep in the stack, from SDP to your cloud bill.

  • WebRTC architecture

    P2P mesh vs. SFU vs. CPaaS decided on your traffic shape, latency budget, and cost ceiling — before you build it twice.

  • Signaling design

    Session setup, renegotiation, reconnection, and presence that survive flaky networks — WebSocket, SIP, or WHIP/WHEP.

  • TURN / STUN & NAT traversal

    coturn deployment, relay-ratio tuning, symmetric-NAT strategy, and ICE configurations that complete in hostile networks.

  • SFU selection & scaling

    Janus, mediasoup, LiveKit, Pion — benchmarked against your load, with simulcast and bandwidth-estimation tuning.

  • Debugging & call-quality forensics

    webrtc-internals, RTCP stats, and packet captures read properly: jitter, loss, freezes, and one-way audio traced to root cause.

  • CPaaS cost & vendor strategy

    Vendor-neutral reviews of Twilio, Daily, Vonage, and LiveKit Cloud bills — and migration paths to hybrid or self-hosted when the math says so.

Simplified WebRTC architecture: two browser peers exchange session setup through a signaling server, fall back to a TURN relay when direct paths fail, and route group media through an SFU. Browser peer A Browser peer B Signaling TURN / STUN SFU
Peers negotiate via signaling, relay through TURN when direct paths fail, and scale group media through an SFU.

JanusmediasoupLiveKitPioncoturnTwilioDailyVonage

Stack-agnostic. I recommend what fits your problem, not what I’m selling.

how we can work together

Three ways in. All of them start with a conversation.

  • Architecture Audit

    from $6,500 · fixed price · 2 weeks

    A full review of your signaling, media topology, TURN layer, and cost structure — delivered as a written report with a prioritized fix list your team can execute.

    Fixed price. Written findings you keep either way.

  • Debugging Sprint

    from $4,500 / week

    Hands-in-the-code, weekly cadence. I work inside your repo and your traces until the failure is found and fixed — connection drops, one-way audio, quality collapse at scale.

    packetsLost / s
    before → after
  • Advisory Retainer

    from $3,000 / month

    Ongoing access for teams shipping real-time features: architecture reviews, incident support, vendor decisions, and a specialist on call when it matters.

Where a fixed scope doesn’t fit, hourly engagements run $225/hr.

what happens next

From first email to work underway in about a week.

  1. Free 30-minute triage

    You describe the failure; I ask the questions that narrow it. No charge, no obligation, and you’ll leave with at least one concrete lead.

  2. Scoped proposal in 48 hours

    A written scope with a fixed price or weekly rate, deliverables, and a timeline — so you know exactly what you’re buying.

  3. Work starts within a week

    Audits begin with your traces and configs; sprints begin in your repo. You see progress in days, not quarters.

photo coming soon

about

I’m Michael. I’ve spent 20+ years making real-time systems work.

Long before “WebRTC” had a name I was shipping systems where latency was the product. Since then I’ve designed signaling layers, deployed and tuned TURN infrastructure, scaled SFU topologies, and debugged the calls other teams had given up on — across telehealth, collaboration, voice AI, and teleoperation workloads.

I work at the protocol level — ICE, SDP, RTP/RTCP, DTLS-SRTP, WHIP/WHEP — and at the invoice level, because most WebRTC problems are ultimately both. When something newer fits better, like WebTransport, I’ll tell you that too.

One consultant, senior only. When we work together, the person you talked to is the person in the code.

Stop guessing why the calls drop.

Tell me what’s failing. The triage call is free, and you’ll know within 30 minutes whether I can fix it.

Book a consultation

Usually scheduled within 48 hours.