Victor is available for hire
Hire VictorVictor Gaydov
Verified Expert in Engineering
Low Latency Developer
Da Nang, Vietnam
Toptal member since July 3, 2026
Bio
Victor builds real-time and low-latency systems: audio/video/voice, networking, synchronization, OS-level work, virtual audio devices. Over 15 years — the last 7 as an independent contractor — he has built production solutions for embedded devices, desktops, phones and cloud.
Portfolio
ViveSound
Audio, Core Audio, Windows Audio Session API (WASAPI)...
Self-employed
Audio, Real-time Streaming, Real-time Audio Processing, Low-latency Software...
TradingView
Go, Streaming Data, High-load, Performance, Networking, Optimization...
Experience
- C++ - 15 years
- Networking - 15 years
- Real-time Streaming - 15 years
- Low Latency - 15 years
- Audio - 10 years
- Go - 10 years
- Embedded Linux - 10 years
- WebRTC - 7 years
Preferred Environment
Audio, C, C++, Go, Real-time Systems, Low Latency, Networking, Linux, Embedded Systems, Real-time Streaming
The most amazing...
...thing I've built is a device that lets musicians in different cities play together over the internet — under 30 ms end-to-end, now manufactured and selling.
Work Experience
Technical Co-founder
2024 - PRESENT
ViveSound
- Joined as a freelance engineer, led it to commercial launch, and became an equity partner.
- Owned the technical roadmap, release cycle, and QA process.
- Designed multi-process architecture spanning background services, virtual devices, and OS integrations.
- Built OS-level audio interception and processing on Windows and macOS (WASAPI, Core Audio).
- Achieved zero overhead with Windows APO, around 3 ms with macOS HAL plugin over shared memory.
- Implemented per-app DSP chains with voice effects, plugin hosting, and cloud voice AI.
Technologies: Audio, Core Audio, Windows Audio Session API (WASAPI), Real-time Audio Processing, Audio Processing, C++, Go, Objective-C, MacOS, Windows, Software Architecture, Real-time Systems, Real-time Computing, Low-latency Software, Audio Engineering, Video & Audio Processing, Audio Engines, Desktop App, Native Mobile Apps, Windows API, Plugins, Multithreading, Concurrency, Performance, Cross-platform, Cross-platform App Development, gRPC, C#, Digital Signal Processing, Git, QML, REST APIs, Architecture, Performance Optimization, Refactoring, Testing, Technical Leadership, Code Review, Operating Systems, Debugging, Test Automation, Large Language Models (LLMs), State Machines, Ubuntu, Docker, JavaScript, Asynchronous Programming, Distributed Applications, Distributed Architecture, SDK Development, SDKs, Amazon S3 (AWS S3), Claude Code, Cloud Storage, Data Migration, Data Reconciliation, Jupyter Notebook, Monorepos, Software Development Toolkits (SDK), Artificial Intelligence (AI), AI Agent Orchestration, LLM Agents, Troubleshooting, iOS, Mobile App Development, Mobile Development, Notifications, Push Notifications, Streaming, Event Streaming, DSP, HAL, Android Development, Speech Recognition, Android, Audio Capture, C++17, Goroutines, Lock-Free Programming, Protobuf, Protocol Design
Real-time Audio & Streaming Engineer
2019 - PRESENT
Self-employed
- Delivered around 10 projects, spanning 7 years, as an independent contractor.
- Implemented probabilistic NAT traversal for P2P streaming (ICE, STUN).
- Built low-latency audio streaming on embedded Linux for a consumer music-collaboration device.
- Built real-time wearable audio augmentation for live venues.
- Built a GPU speech-to-text server with Whisper and WebRTC.
- Implemented an NTP server and client for Zephyr RTOS.
- Built wireless multi-room audio software with under 2 ms synchronization.
- Built synchronized video monitoring for remote operators.
- Developed a real-time software hearing aid for Android and iOS.
- Implemented software for automated voice call testing.
Technologies: Audio, Real-time Streaming, Real-time Audio Processing, Low-latency Software, Core Audio, Windows Audio Session API (WASAPI), Embedded Linux, Embedded Systems, C++, C, Rust, Go, Networking, Sync, Low-level Programming, Bluetooth, Performance, Back-end, Linux, Embedded C, Embedded C++, Real-time Systems, Audio Processing, Audio Engineering, Video & Audio Processing, WebRTC, Opus, RTP, RTSP, FFmpeg, GStreamer, Network Protocols, Network Programming, STUN, TURN (Traversal Using Relay Around NAT), NAT, I/O, Non-blocking I/O, Asynchronous I/O, REST, HTTP, HTTP2, Multithreading, Concurrency, Optimization, Software Architecture, P2P, ALSA, AVFoundation, Android, Android AAudio API, Oboe, iOS, High-load, Git, REST APIs, Architecture, Network Engineering, Performance Optimization, Testing, TCP/IP, gRPC, Distributed Systems, Code Review, Operating Systems, Python, Java, Debugging, Cloud Infrastructure, Test Automation, Large Language Models (LLMs), State Machines, ARM Embedded, Ubuntu, Docker, Asynchronous Programming, Distributed Applications, Distributed Architecture, Encryption, SDK Development, SDKs, Applied Cryptography, Claude Code, Data Migration, Jupyter Notebook, Monorepos, Software Development Toolkits (SDK), Artificial Intelligence (AI), AI Agent Orchestration, LLM Agents, Troubleshooting, NVIDIA CUDA, Native Mobile Apps, Cross-platform, Cross-platform App Development, Mobile App Development, Mobile Development, Flutter, Notifications, Push Notifications, Real-time Tracking, Streaming, DSP, HAL, Android Development, Speech Recognition, Firmware, WiFi, Embedded Development, Android Audio Service, Automatic Speech Recognition (ASR), C++17, Hardware-in-the-loop (HIL), Lock-Free Programming, Memory Pools, Model Deployment, Network Time Protocol (NTP), ntpd, Protobuf, Protocol Design, SIMD, STUN/TURN, System Programming, Time Synchronization, Vectorization, Whisper, Zephyr
Team Lead
2016 - 2019
TradingView
- Developed a high-load streaming hub in Go for trading data (30,000 streams, 20,000 rps, 400 MBps).
- Led a team of 5–10, designed architecture, reviewed code, and owned releases.
- Implemented rate limiting, backpressure, admission control by memory, and contention.
- Debugged bottlenecks down to Go runtime: GC pauses, scheduler run-queue latency.
Technologies: Go, Streaming Data, High-load, Performance, Networking, Optimization, Team Management, Software Architecture, Code Review, Real-time Data, Real-time Data Pipelines, Back-end, Back-end APIs, Concurrency, Multithreading, Performance Measurement, Distributed Systems, Scalable Architecture, Back-end Architecture, Git, Jira, REST APIs, Architecture, Network Engineering, Performance Optimization, Refactoring, Testing, Technical Leadership, TCP/IP, HTTP, Linux, Debugging, Cloud Infrastructure, Test Automation, Ubuntu, PostgreSQL, Docker, Asynchronous Programming, Distributed Applications, Distributed Architecture, Amazon S3 (AWS S3), Cloud Storage, Data Migration, Data Reconciliation, Troubleshooting, Streaming, Event Streaming, Goroutines, Lock-Free Programming
Lead Software Engineer
2014 - 2016
Zodiac Interactive
- Developed a video recording and replay stack for digital-TV set-top boxes running embedded Linux, letting the viewer pause or rewind live TV (Broadcom; ARM, MIPS; uclibc).
- Designed the storage as a filesystem-as-a-database on ext4, with on-disk ring buffers of video chunks.
- Developed an asynchronous caching data layer over four vendor middleware APIs (NDS/Cisco DARWIN).
- Built reproducible cross-builds for 8 set-top platforms.
Technologies: Embedded Linux, Storage, Video Streaming, ARM, MIPS, Performance, Latency & Throughput Analysis, Embedded C, Embedded C++, Embedded Systems, C, C++, Low-level Programming, Performance Measurement, IPTV, Digital TV, Cross-compilation, Git, Confluence, Jira, Architecture, Network Engineering, Performance Optimization, Refactoring, Technical Leadership, TCP/IP, Code Review, Operating Systems, Linux, Debugging, Test Automation, State Machines, ARM Embedded, Ubuntu, Docker, Asynchronous Programming, Encryption, SDK Development, SDKs, Applied Cryptography, Data Migration, Data Reconciliation, Monorepos, Software Development Toolkits (SDK), Troubleshooting, Streaming, Event Streaming, HAL, Python, Embedded Development, System Programming
Software Engineer
2012 - 2014
InfoTeCS, JSC
- Developed a display server for a touchscreen microcontroller, driven over USB (ARM Cortex-M3, no OS).
- Built a serial-over-IP solution, exposing remote serial ports as local device nodes (Linux, CUSE).
- Developed a C++ library for ASN.1 and ruToken/eToken integration.
- Ported a large legacy codebase to the PowerPC architecture.
Technologies: C++, Embedded Linux, PKCS #11, Security, Linux Kernel Programming, PowerPC, Legacy Code, Back-end, Embedded Systems, Network Protocols, Network Programming, ASN.1, Git, Jira, Network Engineering, Performance Optimization, Refactoring, Testing, TCP/IP, HTTP, Code Review, Operating Systems, Linux, Debugging, Test Automation, State Machines, ARM Embedded, Ubuntu, Asynchronous Programming, Distributed Applications, Encryption, SDK Development, SDKs, Applied Cryptography, Data Migration, Data Reconciliation, Monorepos, Software Development Toolkits (SDK), Troubleshooting, Cross-platform, Cross-platform App Development, Streaming, Event Streaming, HAL, Firmware, Serial Communication, Embedded Development, Code Generators, IDL, Libusb, Protocol Design, USB
Software Engineer
2010 - 2012
NTMR
- Developed firmware for an access-control millimeter-wave body scanner (Altera FPGA soft CPU, MicroC/OS-II, C++).
- Built device control over Avalon, UART, CAN, LPT: servo drive, transceiver ADCs, stepper rigs.
- Built the sample acquisition path from scanner to PC over gigabit Ethernet, around 900 Mbit/s UDP (Linux, Windows).
- Implemented distributed classifier training in C++ and MPI on a computing cluster.
Technologies: C, Qt, Qwt, OpenCV, OpenGL, MATLAB, C++, Embedded Systems, FreeRTOS, Real-time Operating System (RTOS), Firmware Development, Firmware, Linux, Windows, Windows API, CAN Bus, Performance Measurement, Scientific Computing, Cluster Computing, Distributed Computing, Digital Signal Processing, Computer Vision, Git, Network Engineering, Performance Optimization, Testing, TCP/IP, gRPC, Distributed Systems, Operating Systems, Debugging, Test Automation, Sensors & Actuators, State Machines, ARM Embedded, Ubuntu, Asynchronous Programming, Distributed Applications, Data Migration, Monorepos, Troubleshooting, NVIDIA CUDA, Cross-platform, Cross-platform App Development, Real-time Tracking, Streaming, Event Streaming, HAL, Serial Communication, Embedded Development, Protocol Design, Technical Leadership
Experience
ViveSound — Desktop Audio Engine (Company Co-founder)
https://vivesound.com/ViveEngine is a desktop audio engine that adds real-time processing to any desktop app, without the app knowing.
I joined as a freelance engineer, led the product to commercial launch, and became an equity partner. Today, I own the roadmap, release cycle, and QA. Built the multi-process architecture, OS-level audio interception with WASAPI and Core Audio, per-app effect chains with voice effects, and plugin hosting.
It has zero overhead with the Windows APO, and around 3 ms with the macOS HAL plugin over shared-memory transport. It was shipped as a redistributable engine on Windows 10/11 and macOS, signed and notarized.
Platforms: Windows, macOS
I joined as a freelance engineer, led the product to commercial launch, and became an equity partner. Today, I own the roadmap, release cycle, and QA. Built the multi-process architecture, OS-level audio interception with WASAPI and Core Audio, per-app effect chains with voice effects, and plugin hosting.
It has zero overhead with the Windows APO, and around 3 ms with the macOS HAL plugin over shared-memory transport. It was shipped as a redistributable engine on Windows 10/11 and macOS, signed and notarized.
Platforms: Windows, macOS
Roc Streaming — Real-time Audio over IP (Open Source)
https://roc-streaming.org/Roc Streaming is an open-source ecosystem for real-time audio over IP, over Wi-Fi, cable, or the Internet.
I co-founded the project and remain one of its core maintainers. I designed and implemented key parts of the streaming stack: adaptive jitter buffer, packet-loss recovery with Reed-Solomon and LDPC FEC, clock synchronization, and 11 network protocol RFCs. I co-authored the surrounding ecosystem: PulseAudio modules, a macOS virtual device, an Android app, and Java and Go bindings.
It's down to 10 ms end-to-end, with 1-2 ms synchronization error between receivers. It's used by the PipeWire team as one of its networking options. It also has 1,200 stars and 50+ contributors.
Platforms: Linux, macOS, Android; x86_64, i686, ARM64, ARM32, RISC-V, MIPS32
I co-founded the project and remain one of its core maintainers. I designed and implemented key parts of the streaming stack: adaptive jitter buffer, packet-loss recovery with Reed-Solomon and LDPC FEC, clock synchronization, and 11 network protocol RFCs. I co-authored the surrounding ecosystem: PulseAudio modules, a macOS virtual device, an Android app, and Java and Go bindings.
It's down to 10 ms end-to-end, with 1-2 ms synchronization error between receivers. It's used by the PipeWire team as one of its networking options. It also has 1,200 stars and 50+ contributors.
Platforms: Linux, macOS, Android; x86_64, i686, ARM64, ARM32, RISC-V, MIPS32
libASPL — C++ Library for macOS Virtual Devices (Open Source)
https://github.com/gavv/libASPLlibASPL is an open-source C++17 library for building virtual audio devices on macOS.
I built a C++ layer over the Core Audio HAL, with real-time-safe audio I/O based on lock-free data structures. I also wrote a Core Audio HAL simulation for driver testing.
It's used in production by open-source and commercial projects.
Platforms: macOS
I built a C++ layer over the Core Audio HAL, with real-time-safe audio I/O based on lock-free data structures. I also wrote a Core Audio HAL simulation for driver testing.
It's used in production by open-source and commercial projects.
Platforms: macOS
Consumer Audio Device for Online Music Collaboration (Freelance Project)
A hardware box that lets musicians play together over the Internet with live instruments.
I built the audio pipeline on embedded Linux and bare ALSA on a custom PCB: real-time path clocked by ALSA, with lock-free data structures, arena allocators, and NEON on hot paths. There's an adaptive jitter buffer and FEC for packet loss. I drove the device from prototype to production over two years.
Under 30 ms end-to-end, ADC to DAC over the Internet, on a direct peer-to-peer path. It's manufactured and selling, handed over to in-house engineers.
Platforms: embedded Linux; ARM64 (Cortex-A72)
I built the audio pipeline on embedded Linux and bare ALSA on a custom PCB: real-time path clocked by ALSA, with lock-free data structures, arena allocators, and NEON on hot paths. There's an adaptive jitter buffer and FEC for packet loss. I drove the device from prototype to production over two years.
Under 30 ms end-to-end, ADC to DAC over the Internet, on a direct peer-to-peer path. It's manufactured and selling, handed over to in-house engineers.
Platforms: embedded Linux; ARM64 (Cortex-A72)
Wearable Audio Augmentation for Live Venues (Freelance Project)
A wireless box that plays a headphone feed aligned with the live sound reaching the listener through the air.
I built the receiver on embedded Linux: audio arrives over Wi-Fi, and alignment is anchored to ALSA hardware timestamps at ADC input and DAC output.
Under 0.25 ms alignment error, no matter the distance from the stage. Measured and calibrated on a hardware test rig.
Platforms: embedded Linux; ARM64 (Cortex-A53)
I built the receiver on embedded Linux: audio arrives over Wi-Fi, and alignment is anchored to ALSA hardware timestamps at ADC input and DAC output.
Under 0.25 ms alignment error, no matter the distance from the stage. Measured and calibrated on a hardware test rig.
Platforms: embedded Linux; ARM64 (Cortex-A53)
Synchronized Video Monitoring for Remote Operators (Freelance Project)
Live video from RTSP cameras fanned out to operators at distant sites who must react in lockstep.
I built the gateway that transcodes the camera streams into RTP/H.264 with KCP FEC. Receivers run an NTP-like clock sync protocol to schedule frame presentation on a shared timeline.
Under 0.5 ms presentation skew between sites.
Platforms: Linux
I built the gateway that transcodes the camera streams into RTP/H.264 with KCP FEC. Receivers run an NTP-like clock sync protocol to schedule frame presentation on a shared timeline.
Under 0.5 ms presentation skew between sites.
Platforms: Linux
Probabilistic NAT Traversal (Freelance Project)
A layer on top of ICE that widens the range of NATs where two peers can connect directly, without a relay in the middle.
I wrote it in C on embedded Linux, around libnice: classify the NAT, open the hole, and hand back to ICE. Classification is a custom STUN client doing RFC 5780 behavior discovery. Traversal is probabilistic — one side holds 250 ports open while the other sends random probes, so the birthday paradox does the work.
Direct P2P between symmetric and port-restricted cone NATs, where stock ICE gives up and falls back to a relay. One probe batch connects about 85% of the time, and the connection comes up in 1-2 minutes. Running in production, tested against real ISPs.
Platforms: embedded Linux; ARM64
I wrote it in C on embedded Linux, around libnice: classify the NAT, open the hole, and hand back to ICE. Classification is a custom STUN client doing RFC 5780 behavior discovery. Traversal is probabilistic — one side holds 250 ports open while the other sends random probes, so the birthday paradox does the work.
Direct P2P between symmetric and port-restricted cone NATs, where stock ICE gives up and falls back to a relay. One probe batch connects about 85% of the time, and the connection comes up in 1-2 minutes. Running in production, tested against real ISPs.
Platforms: embedded Linux; ARM64
Automated Voice Call Testing Over Bluetooth (Freelance Project)
A lab that scores voice call quality on real phones, with no human holding them.
The lab machine pairs with each phone as a Bluetooth HFP headset, the only way into call audio on a stock phone. It captures and injects call audio and controls the phones. Live monitoring runs in the browser over WebRTC, with recordings uploaded to the cloud for analysis.
Scores calls with PESQ/MOS. Two adapters let two phones call each other through the lab, so both ends are scored at once.
Platforms: Linux
The lab machine pairs with each phone as a Bluetooth HFP headset, the only way into call audio on a stock phone. It captures and injects call audio and controls the phones. Live monitoring runs in the browser over WebRTC, with recordings uploaded to the cloud for analysis.
Scores calls with PESQ/MOS. Two adapters let two phones call each other through the lab, so both ends are scored at once.
Platforms: Linux
Software Hearing Aid for Android and iOS (Freelance Project)
Turns a phone into a hearing aid: it captures the room through the mic, applies the user's hearing profile, and plays the result to the headset.
I built one portable C++ core with thin native shells: AAudio/Oboe on Android, Core Audio/AVFoundation on iOS. The real-time path joins input and output callbacks through a lock-free ring buffer, with configuration swapped via a double buffer.
50 ms mouth-to-ear on budget phones from 2020 (Redmi Note 7 — 48 ms, iPhone X — 50 ms), measured acoustically on a purpose-built rig. QA on 9 devices picked by market share.
Platforms: Android, iOS
I built one portable C++ core with thin native shells: AAudio/Oboe on Android, Core Audio/AVFoundation on iOS. The real-time path joins input and output callbacks through a lock-free ring buffer, with configuration swapped via a double buffer.
50 ms mouth-to-ear on budget phones from 2020 (Redmi Note 7 — 48 ms, iPhone X — 50 ms), measured acoustically on a purpose-built rig. QA on 9 devices picked by market share.
Platforms: Android, iOS
GPU Speech-to-text Server with WebRTC and Whisper (Freelance Project)
A self-hosted server for real-time transcription: browser audio in over WebRTC, transcript out.
I built the whole path: browser capture, Go/Pion transport, faster-whisper over CTranslate2 on the GPU. Whisper is not a streaming model, so a sliding window is re-transcribed as it moves, retiring settled speech.
It was shipped and deployed to the customer's AWS cloud.
Platforms: Linux
I built the whole path: browser capture, Go/Pion transport, faster-whisper over CTranslate2 on the GPU. Whisper is not a streaming model, so a sliding window is re-transcribed as it moves, retiring settled speech.
It was shipped and deployed to the customer's AWS cloud.
Platforms: Linux
Wireless Audio Multiroom Software (Freelance Project)
Wireless speakers in different rooms, playing the same audio in sync.
I built the synchronization layer: clock offset is estimated by NTP-like exchanges, filtered to reject Wi-Fi delay spikes. Drift is corrected against measured DAC playout time by a PI-driven resampler.
Under 2 ms synchronization between speakers, measured on a hardware test rig.
Platforms: Linux, macOS
I built the synchronization layer: clock offset is estimated by NTP-like exchanges, filtered to reject Wi-Fi delay spikes. Drift is corrected against measured DAC playout time by a PI-driven resampler.
Under 2 ms synchronization between speakers, measured on a hardware test rig.
Platforms: Linux, macOS
Embedded Video Storage with Recording and Replay (Zodiac Interactive)
A DVR stack for digital-TV set-top boxes: live TV is recorded continuously, and the viewer can pause or rewind at any moment.
I served as the lead developer of the two C++ libraries behind it. Storage is a filesystem-as-a-database on ext4, with on-disk ring buffers of video chunks. Pause and catch-up are hardlinks over the same chunks, with no data copy. A long-running synthetic stress test ran the code on the host to reproduce field faults.
It was shipped on 8 set-top platforms, on Broadcom SoCs.
Platforms: embedded Linux (uclibc); MIPS, ARM
I served as the lead developer of the two C++ libraries behind it. Storage is a filesystem-as-a-database on ext4, with on-disk ring buffers of video chunks. Pause and catch-up are hardlinks over the same chunks, with no data copy. A long-running synthetic stress test ran the code on the host to reproduce field faults.
It was shipped on 8 set-top platforms, on Broadcom SoCs.
Platforms: embedded Linux (uclibc); MIPS, ARM
NTP Server and Client for Zephyr RTOS (Freelance Project)
A time server running on an STM32 board, for machines on an isolated network segment.
I ported the NTP server from BusyBox to Zephyr, reduced to single-peer mode. Wrote an SNTP client for Zephyr from scratch: unlike the built-in one, it corrects for network delay as per RFC 2030 and supports a hardware RTC.
It was delivered with two multi-machine testbeds: real hardware for QA, QEMU for development.
Platforms: Zephyr RTOS; ARM Cortex-M7 (STM32F7)
I ported the NTP server from BusyBox to Zephyr, reduced to single-peer mode. Wrote an SNTP client for Zephyr from scratch: unlike the built-in one, it corrects for network delay as per RFC 2030 and supports a hardware RTC.
It was delivered with two multi-machine testbeds: real hardware for QA, QEMU for development.
Platforms: Zephyr RTOS; ARM Cortex-M7 (STM32F7)
Microcontroller Display Server Over USB (InfoTeCS, JSC)
The front-panel touchscreen console of a rack-mounted HSM, running on a microcontroller with no OS.
I wrote the firmware as a display server: the MCU handles drawing primitives, hit-testing, and multi-tap, while the Linux host builds widgets on top over a 62 KB/s USB HID link. An SDL emulator runs the same firmware code on the desktop, through an alternative HAL implementation.
Console flows are described in XML and generated into C++, so the product team could build around 60 wizards without firmware work or a board on the desk.
Platforms: bare-metal ARM Cortex-M3 at 80 MHz, Linux host
I wrote the firmware as a display server: the MCU handles drawing primitives, hit-testing, and multi-tap, while the Linux host builds widgets on top over a 62 KB/s USB HID link. An SDL emulator runs the same firmware code on the desktop, through an alternative HAL implementation.
Console flows are described in XML and generated into C++, so the product team could build around 60 wizards without firmware work or a board on the desk.
Platforms: bare-metal ARM Cortex-M3 at 80 MHz, Linux host
Serial-over-IP for Remote Industrial Equipment (InfoTeCS, JSC)
Industrial equipment on RS-232 or RS-485 in a factory appears as a device node in /dev on a workstation in the office.
The factory runs a server that owns the real port and exposes a protocol covering termios concepts like parity, baud, and control lines. The workstation runs a client that creates a virtual character device via CUSE and redirects I/O and ioctls to the remote host.
The client's third-party software uses the remote port as a regular serial port.
Platforms: Linux
The factory runs a server that owns the real port and exposes a protocol covering termios concepts like parity, baud, and control lines. The workstation runs a client that creates a virtual character device via CUSE and redirects I/O and ioctls to the remote host.
The client's third-party software uses the remote port as a regular serial port.
Platforms: Linux
Firmware for Millimeter-wave Body Scanner (NTMR, LLC)
An access-control body scanner: antennas ride a rotating frame around a person and sweep a microwave band.
I wrote the board firmware on MicroC/OS-II, running on a soft CPU inside an Altera FPGA: task framework plus drivers for the servo drive over UART, stepper rigs over CAN, and the encoder, transceiver ADCs, and flash controller over the Avalon bus.
Raw samples stream from the FPGA to the PC over UDP at around 900 Mbit/s.
Platforms: MicroC/OS-II RTOS on an Altera FPGA soft CPU
I wrote the board firmware on MicroC/OS-II, running on a soft CPU inside an Altera FPGA: task framework plus drivers for the servo drive over UART, stepper rigs over CAN, and the encoder, transceiver ADCs, and flash controller over the Avalon bus.
Raw samples stream from the FPGA to the PC over UDP at around 900 Mbit/s.
Platforms: MicroC/OS-II RTOS on an Altera FPGA soft CPU
Roc VAD — Virtual Audio Device For macOS (Open Source)
https://github.com/roc-streaming/roc-vadRoc VAD adds virtual speakers and microphones to macOS that stream audio over the network.
I built a userspace Core Audio HAL plug-in. The network pipeline runs inside the sound server, so there is no IPC on the audio path.
Controllable programmatically over gRPC, and vendorable into third-party products.
Platforms: macOS
I built a userspace Core Audio HAL plug-in. The network pipeline runs inside the sound server, so there is no IPC on the audio path.
Controllable programmatically over gRPC, and vendorable into third-party products.
Platforms: macOS
Roc Droid — Wi-Fi Audio with System Capture for Android (Open Source)
https://github.com/roc-streaming/roc-droidRoc Droid turns an Android phone into a wireless speaker, or lets it play through another device.
I co-authored the app: system-wide capture of what other apps play, via MediaProjection, plus streaming to and from remote endpoints. I used Flutter, Kotlin, JNI, and C++.
It interoperates with PipeWire, macOS, and embedded endpoints of the Roc ecosystem. It was published on the F-Droid app store.
Platforms: Android
I co-authored the app: system-wide capture of what other apps play, via MediaProjection, plus streaming to and from remote endpoints. I used Flutter, Kotlin, JNI, and C++.
It interoperates with PipeWire, macOS, and embedded endpoints of the Roc ecosystem. It was published on the F-Droid app store.
Platforms: Android
webrtc-cli — WebRTC Bridge Between Sound Cards and Browser (Open Source)
https://github.com/gavv/webrtc-cliwebrtc-cli is an open-source command-line tool that streams audio between local sound cards or files and a browser, over WebRTC.
I wrote it in Go with Pion, with PulseAudio and Opus support, including a custom Opus depacketizer with in-band FEC recovery and PLC fallback, an adaptive jitter buffer, and simplistic clock drift handling.
It's open-sourced from a freelance project, where it runs in production.
Platforms: Linux
I wrote it in Go with Pion, with PulseAudio and Opus support, including a custom Opus depacketizer with in-band FEC recovery and PLC fallback, an adaptive jitter buffer, and simplistic clock drift handling.
It's open-sourced from a freelance project, where it runs in production.
Platforms: Linux
Signal Estimator — Latency Measurement Tool (Open Source)
https://github.com/gavv/signal-estimatorSignal Estimator is an open-source tool for measuring latency and other properties of an audio path.
It plays a probe signal, captures it looped back, and computes end-to-end delay, I/O jitter, and synchronicity of multiple inputs. ALSA driver timestamps split the measured latency into hardware and software parts.
I use it to build software and hardware test benches: local sound cards, network streaming, virtual devices, and acoustic paths. Most of the figures in this portfolio were measured with it.
Platforms: Linux; x86_64, ARM64, ARM32
It plays a probe signal, captures it looped back, and computes end-to-end delay, I/O jitter, and synchronicity of multiple inputs. ALSA driver timestamps split the measured latency into hardware and software parts.
I use it to build software and hardware test benches: local sound cards, network streaming, virtual devices, and acoustic paths. Most of the figures in this portfolio were measured with it.
Platforms: Linux; x86_64, ARM64, ARM32
Skills
Libraries/APIs
Android AAudio API, WebRTC, FFmpeg, Core Audio, Windows API, Pion, REST APIs, Protobuf, Libusb, Back-end APIs, Qwt, OpenCV, OpenGL, PKCS #11
Tools
TURN (Traversal Using Relay Around NAT), Git, Claude Code, Hardware-in-the-loop (HIL), ntpd, Time Synchronization, Whisper, Zephyr, MATLAB, Confluence, Jira
Languages
Go, C++, C, Embedded C, Embedded C++, IDL, MIPS, C++17, Python, Rust, Dart, C#, Java, Objective-C, QML, Swift, Kotlin, JavaScript
Frameworks
gRPC, JNI, GStreamer, Flutter, Qt, PowerPC
Paradigms
Real-time Systems, Cross-platform, REST, Refactoring, Testing, Asynchronous Programming, Mobile Development, RISC-V, Back-end Architecture, Test Automation, ASN.1, Distributed Computing
Platforms
Linux, MacOS, Windows, Embedded Linux, Ubuntu, iOS, Android, Docker, Jupyter Notebook, NVIDIA CUDA
Storage
Sync, PostgreSQL, Amazon S3 (AWS S3)
Other
PulseAudio, PipeWire, AVFoundation, Windows Audio Session API (WASAPI), RTSP, Opus, System Programming, Networking, Multithreading, Concurrency, TCP/IP, Audio, Real-time Streaming, Low-latency Software, Real-time Audio Processing, Audio Processing, Software Architecture, Embedded Systems, Streaming Data, High-load, Performance, Team Management, Low-level Programming, Audio Streaming, Low Latency, ALSA, Network Programming, Network Protocols, STUN, NAT, Real-time Computing, Audio Engineering, Video & Audio Processing, Audio Engines, Native Mobile Apps, Cross-platform App Development, Back-end, RTP, I/O, Non-blocking I/O, Asynchronous I/O, HTTP, P2P, Oboe, RTCP, SDP, ICE, Performance Measurement, Code Review, Real-time Data, Real-time Data Pipelines, Scalable Architecture, Firmware Development, Architecture, Network Engineering, Performance Optimization, Technical Leadership, Operating Systems, Debugging, State Machines, ARM Embedded, Distributed Applications, Distributed Architecture, SDK Development, SDKs, Monorepos, Software Development Toolkits (SDK), Troubleshooting, Mobile App Development, Notifications, Push Notifications, Streaming, Event Streaming, HAL, Embedded Development, Android Audio Service, Lock-Free Programming, Memory Pools, Network Time Protocol (NTP), Protocol Design, SIMD, STUN/TURN, Vectorization, Code Generators, USB, Audio Capture, Goroutines, Open Source, Performance Testing, ARM, Optimization, Storage, Video Streaming, Latency & Throughput Analysis, Legacy Code, Digital Signal Processing, HTTP2, Distributed Systems, Signal Analysis, Cross-compilation, Real-time Operating System (RTOS), Firmware, BlueZ, Large Language Models (LLMs), Sensors & Actuators, Data Migration, Data Reconciliation, Artificial Intelligence (AI), LLM Agents, Real-time Tracking, DSP, Android Development, Speech Recognition, WiFi, Serial Communication, Automatic Speech Recognition (ASR), Model Deployment, H.264, H264, Video Codecs, Security, Linux Kernel Programming, FreeRTOS, Bluetooth, Desktop App, Plugins, IPTV, Digital TV, CAN Bus, Scientific Computing, Cluster Computing, Computer Vision, Yocto, Cloud Infrastructure, Encryption, Applied Cryptography, Cloud Storage, AI Agent Orchestration
Collaboration That Works
How to Work with Toptal
Toptal matches you directly with global industry experts from our network in hours—not weeks or months.
1
Share your needs
Discuss your requirements and refine your scope in a call with a Toptal domain expert.
2
Choose your talent
Get a short list of expertly matched talent within 24 hours to review, interview, and choose from.
3
Start your risk-free talent trial
Work with your chosen talent on a trial basis for up to two weeks. Pay only if you decide to hire them.
Top talent is in high demand.
Start hiring