Voice AI engineering · Ifeanyi Dike

Make your voice agent work beyond the demo.

I help teams build voice agents, connect them to their systems, and fix the problems that show up in real calls.

Tell me what you’re building See my engineering work

Based in Port Harcourt, Nigeria. Working remotely.

Inside a voice agentA typical call flow
  1. 01
    ListenPhone audio → speech to text
  2. 02
    RespondModel + your tools and data
  3. 03
    SpeakText to speech → caller

Timing, interruptions and failures matter at every step.

PythonVapiPipecatLiveKitTwilioFastAPIPostgreSQLRedis

How I can help

From the first call to the hard-to-find bug.

Build and connect

Build and debug Vapi integrations, connect custom backends, or migrate to Pipecat when you need more control.

Find what’s breaking

Trace dead air, missed speech, broken handoffs and recording errors. Find the cause and check the fix.

Prepare for real use

Improve call cleanup, queues, retries and resource use. Add the logs your team needs to understand a failed call.

Selected engineering work

The work behind the offer.

Examples from my engineering roles, explained in public write-ups.

Voice infrastructureFrom managed platform to Pipecat

Taking control of the call pipeline

I migrated a production voice system to Pipecat, connecting Twilio, speech services and a FastAPI backend. The work included call logs, turn-taking and safe cleanup when calls failed.

Read the migration notes
Call debuggingRecordings that hid the real problem

When the recording tells a different story

A recording could contain words that never reached the caller. I traced the capture and timing problems, then moved to provider recordings with separate channels and a shared clock.

Read the investigation
70%+Lower memory use across two production Python services

Fixing the cause of memory-limit restarts

I used Memray to find repeated model-tokenizer allocations, changed how workers shared memory, and removed an unnecessary import from a background worker.

Read the profiling notes

Working together

Start with a clear, small scope.

You can bring me in for one problem or a defined part of a build. Agencies are welcome too.

  1. Tell me the problem

    Share your current setup, what goes wrong and what you need to change.

  2. Agree on the work

    We set the scope, fee, delivery date and how we’ll check the result.

  3. Get a useful handover

    You get the agreed work, clear notes and the steps your team needs next.

Ifeanyi Dike

The engineer you’ll work with

Hi, I’m Ifeanyi.

I’m a software engineer and applied mathematician with over eight years in software. At DataStrut, I work on production voice AI and the backend systems around it. I also write about the problems I solve and research how speech agents are evaluated.

My experience GitHub Engineering blog

Have a voice AI problem?

Let’s look at what you need.

Send a short note about your stack, the problem and the outcome you want. Please leave out credentials and private customer records.

Email me ifeanyi@ifeanyidike.com