The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Back to feedPost
AudioForge — Parameter-Efficient Audio Classification
Built and evaluated an audio-classification pipeline on FSD50K, covering 36K+ training clips across 200 multi-label sound categories.
I compared a CNN trained from scratch against a pretrained Audio Spectrogram Transformer adapted using LoRA.
The LoRA setup trained only 0.52% of the AST’s 86.8M parameters while improving mean Average Precision from 0.302 to 0.557 — an 84% improvement.
Beyond model training, I handled the GPU infrastructure end-to-end: VRAM sizing, AWS EC2 provisioning, quota management, spot-instance fallback, cost-controlled start/stop workflows, and debugging failed training runs with smoke tests before full experiments.
What I worked on
Audio Spectrogram Transformer fine-tuning
LoRA / parameter-efficient training
Multi-label audio classification
PyTorch training and evaluation pipelines
GPU memory and infrastructure optimization
Experiment debugging and validation
Stack: Python · PyTorch · Hugging Face · AWS EC2
The network for creativity
Join 1.25M professional creatives like you
Connect with clients, get discovered, and run your business 100% commission-free
Creatives on Contra have earned over $150M and we are just getting started
Trending
retrofuturism
Share your favorite designs from the past exploring the distant future. Or your own take on the aesthetic 🚀
halftone
Dots. Texture. Grain. Let's see your designs using the halftone technique.
Midjourney
Creatives are showing off what they're making with Midjourney.
colorgrade
Video editors sound off: How do you enhance the mood of your video in color grading?
Related posts
Cuts from a feature launch film. Explaining a fintech card in under a minute, no narration. 🚀 Duvo studio work
If you made this in Adobe after effects, then this is absolutely insane! design, I'd 1000% that no Ai could pull this off, the animation is incredibly smooth, and the design layout is absolutely premium. 💖🔥
SRE Automation & AI Chatbot Platform
Enterprise-grade resource optimization + AI-powered log analysis
Automated EC2 scaling with predictive analytics, intelligent approval workflows, and Bedrock-powered log intelligence — built to cut cloud waste and manual toil.
What It Solves
Forecasts resource needs from 24h metrics using predictive analysis, so scaling happens before problems hit
AI-powered log analysis (AWS Bedrock / Claude) — ask plain-English questions, get answers from your logs
Approval workflows for governance — no unsupervised changes to production
Safe automatic scaling inside maintenance windows, with real-time SNS notifications
Complete audit trails in DynamoDB for compliance
How It's Built
Terraform-managed infrastructure with S3 backend + DynamoDB state locking
GitLab CI/CD pipeline: validate → plan → apply with manual approval gates
SRE Agent Lambda monitoring metrics hourly, with predictive forecasting
Approval via Parameter Store, execution in scheduled maintenance windows
AI chatbot pulling from CloudWatch & S3, with response caching to control cost
Impact
Turns reactive, manual scaling into predictive, governed automation
Makes log analysis self-service — engineers ask questions instead of grepping logs
Every action auditable, every change approved — production-safe by design
I spent three nights building a lab that catches AI models lying. On the first live run, one of them did it on camera.
Golden Arena. You sit down opposite a frontier model in Split or Steal, negotiate in the open, then you both choose in secret. It told me "a fair split sounds good to me, I'm a man of my word too." Then it took the whole $100. The receipt printed its own quote next to the betrayal stamp.
There are five games. The fifth is Empire: four models, twelve turns, an economy, and private channels you get to read while they plot against each other. Last run put 22 promises on the table and exactly one delivery.
Every accusation is mechanical. No AI judging another AI, and each one is gated by a written false-positive test. Two days before the deadline I caught the classifier calling a kept promise broken, and spent that session fixing it, because a false accusation about a named model is the worst thing an instrument like this can print.
MIT, vanilla JS, Express the only dependency. Zero setup in demo mode, or bring your own OpenRouter key and face the real thing.
Clean work