Sound design · full redesign
Battlefield 6 sound redesign
A full sound redesign of footage from EA’s Battlefield 6. The gameplay runs on top with my session underneath, so you can watch each sound land.
Audio engineer · Producer · Speech AI
Graham
Armstrong
I make records, run sessions for Grammy-nominated and major-label artists, and work on the data behind speech AI.
Los Angeles · San Francisco
Music
Records & live sets
Original releases with millions of streams, production for charting pop and R&B acts, contest wins and DJ sets.
Explore music
AI audio
Speech & voice AI
Speech data at ByteDance / TikTok, and a voice model built from two singers.
Explore AI audio
Engineering
In the studio
Sessions for Grammy-nominated and major-label artists, the signal chain I own, and the audio tools I build.
Explore engineering
Music
Records &
live sets
I’m a producer and DJ in Los Angeles. My original releases have millions of streams, and I’ve produced for charting and award-winning pop and R&B acts.
Music biography
From Berklee to Los Angeles
Graham Armstrong is a music producer and DJ based in Los Angeles. His original releases have generated millions of streams on major platforms, and he has produced for charting and award-winning pop and R&B acts. He has won production competitions hosted by Cymatics, Kenny Beats and Producergrind, and he DJs at high-end Los Angeles venues.
He studied music production and audio engineering at Berklee College of Music on a Worldwide Scholarship, then moved to the University of Southern California, where he studies the same field and is on the Dean’s List. Alongside his own records, he engineers sessions for Grammy-nominated and major-label artists, has scored a video for a Spitfire Audio contest, and has redesigned all the sound in footage from EA’s Battlefield 6.
Contest wins
Cymatics · Kenny Beats · ProducergrindProduction competitionsReleases
Millions of streamsOriginal music on major platformsLive
DJ at high-end LA venuesLos AngelesTraining
Berklee → USCWorldwide Scholarship, then Dean’s ListRecent productions
- Cypher
- Overtime
- LightsDown
- 9Deep
Inside a session · TikTok
C’est La Vie
For fun, I post screen recordings of my FL Studio sessions on TikTok. This one is “C’est La Vie”: the sample up top, then TRITON Extreme, Serum 2, ZENOLOGY, Diva and M1, with the bass and drums underneath.
Scoring and sound design
Rap orchestral scoring projects
Rap remixes for scoring practice: I take something that usually isn’t a score and add strings to it.
01
Rap orchestral scoring project
02
Rap orchestral scoring project
Scoring · contest entry
Spitfire Audio scoring contest
My score for a video from a contest held by Spitfire Audio.
AI audio
Speech &
voice AI
At ByteDance / TikTok I owned data quality for multilingual voice datasets in 100+ languages and dialects, used to train speech foundation models. I also build voice models and test how they sound.
Speech data
Speech data at ByteDance / TikTok
Project Engineer Intern, Data Speech Foundation Generation · San Jose, CA · summer 2026
- Owned data quality for multilingual voice datasets spanning 100+ languages and dialects, from raw capture to training-ready delivery for speech foundation models.
- Evaluated 1,000+ hours of speech audio for signal consistency, articulation accuracy, background noise floor and technical compliance.
- Built and maintained transcription-alignment, metadata-verification and dataset-standardization workflows across deliveries from internal teams and external vendors.
- Managed vendor partners and shared data assets across concurrent speech and audio projects supporting TikTok and CapCut teams.
- 100+ languages and dialects
- 1,000+ hours evaluated
- TikTok · CapCut
- Raw captureinternal teams and vendors
- Validationsignal, noise floor, compliance
- Alignment & metadatatranscripts and labels
- Standardizationone format per delivery
- Training-ready deliveryspeech foundation models
Featured · AI in production
Two singers, one new voice
Henry Cornell and Jettrin are both singers in my class at USC. For a class project, I wanted to use both of their unique voices but have the song sung by one voice, so I created a model of both of their voices and made a combined voice. Watch the story, then switch between the originals, the model and the full song yourself.
A class project at USC
Two singers. One new voice.
Henry Cornell and Jettrin, combined into one voice with a model I built. About 50 seconds, with sound.
Two singers. One new voice.
Henry Cornell and Jettrin, classmates at USC. Both agreed to have their voices modeled for this project. Recording and voice model by Graham Armstrong.
Loading audio…
0:00 / 0:51
Hear
Speech AI toolkit
- Multilingual dataset curation
- Speech-data QA
- Transcription alignment
- Metadata validation
- Labeling schema
- ASR training-data quality
- Vendor and data-asset management
Audio engineering
In the
studio
I’m one of the lead recording engineers at IcecreamSound Studios in Los Angeles, running sessions end to end for Grammy-nominated and major-label artists.
Recording
Studios
Los Angeles
- Hip-hop
- R&B
- Pop
- Alternative
IcecreamSound Studios · Recording engineer
- One of the studio’s lead recording engineers, running sessions end to end for Grammy-nominated and major-label artists.
- Built an independent client base through repeat bookings and artist referrals.
- Troubleshoot Pro Tools session faults and signal-flow issues in real time while managing client intake, scheduling and session turnover.
Boston
- Analog outboard
- Headphone mixes
CyberSound Recording Studios · Assistant recording engineer
- Supported and independently ran professional sessions for Grammy-nominated and major-label-affiliated artists.
- Managed signal flow, gain staging, headphone mixes and analog outboard in a high-volume room.
- Handled client scheduling, transactions and studio readiness.
What I own in a session
On the sessions I run, these four are mine, from the first gain stage to the files the artist leaves with.
- Signal chainBuilt for the source and the session
- Gain structureClean levels at every stage
- RoutingInputs, buses and headphone mixes
- Final deliveryFiles checked, named and delivered
Code · DSP · apps
Tools I build
Python · FFmpeg · DSP
Audio batch processing & validation toolkit
Command-line tools that batch-process large multi-session libraries: format conversion, filename and format validation, metadata read and write, and automated integrity checks. They replaced repeated manual passes before client delivery.
Session prep, naming conventions and file delivery became reusable commands.
DSP · Python · audio analysis
Audio plug-ins & signal processing
Experimental audio plug-ins and DSP prototypes in Python, covering production effects, parameter automation and audio analysis.
React Native · TypeScript · Expo
GoFit: workout tracking & progressive overload
A cross-platform mobile app I shipped. Its progressive-overload engine sets prescribed loads from workout history, rep counts, estimated strength and set-speed feedback.
It also handles exercise logging, PR tracking and performance charts.
Toolkit
Skills
Audio & DSP
- Batch audio processing
- Sample-rate and bit-depth validation
- Spectral and level analysis
- Noise-floor assessment
- Gain structure
- Signal-chain design
- Critical listening
DAWs & production
- Pro Tools
- Ableton Live
- FL Studio
- Adobe Audition
- Premiere Pro
- After Effects
Tools & libraries
- FFmpeg
- NumPy
- React Native
- Expo
- Git
- Visual Studio Code
Languages
- Python
- TypeScript
- JavaScript