acaciabengo
Joined 1 year ago
Karma
3
Submissions
15
Member since
Oct 21
Recent submissions
Comment · 1 month ago
Location: Kampala, Uganda (GMT +3)
Remote: Yes
Willing to relocate: No
Technologies: PyTorch, computer vision, object detection, image segmentation, YOLO, NLP, LLMs, multimodal embeddings, FastAPI, GCP, RunPod
Résumé/CV: https://drive.google.com/file/d/1E0xkf_EAjlp12xZpjjXPJaTCA9g...
Email: [email protected]
Available as a hands-on ML/AI contractor.
I build and evaluate production ML systems across computer vision, NLP, and multimodal AI.
Computer vision: detection, segmentation, document understanding, and 3D reconstruction. Most recently, I built a production pipeline that takes floor-plan PDFs and produces wall, door, and window quantities plus 3D geometry at real-world scale, running on serverless GPU infrastructure.
NLP / LLMs: text classification, semantic search, embeddings, content moderation, information extraction, ranking, and automated triage. I've built systems that process large volumes of user-generated content and identify which items actually require human review.
Multimodal AI: combining text, image, and metadata signals for classification, retrieval, evaluation, and document understanding.
ML evaluation: I investigate cases where models look better than they really are, including data leakage. I can work on an existing model or take a problem from prototype through to a production API.
Happy to start with a paid diagnostic using data you already have.
Comment · 1 month ago
Location: Kampala, Uganda
Remote: Yes
Willing to relocate: No
Technologies: PyTorch, computer vision, object detection, image segmentation, YOLO, NLP, LLMs, multimodal embeddings, FastAPI, GCP, RunPod
Résumé/CV: https://drive.google.com/file/d/1E0xkf_EAjlp12xZpjjXPJaTCA9g...
Email: [email protected]
Available as a hands-on ML/AI contractor.
I build and evaluate production ML systems across computer vision, NLP, and multimodal AI.
Computer vision: detection, segmentation, document understanding, and 3D reconstruction. Most recently, I built a production pipeline that takes floor-plan PDFs and produces wall, door, and window quantities plus 3D geometry at real-world scale, running on serverless GPU infrastructure.
NLP / LLMs: text classification, semantic search, embeddings, content moderation, information extraction, ranking, and automated triage. I've built systems that process large volumes of user-generated content and identify which items actually require human review.
Multimodal AI: combining text, image, and metadata signals for classification, retrieval, evaluation, and document understanding.
ML evaluation: I investigate cases where models look better than they really are, including data leakage. I can work on an existing model or take a problem from prototype through to a production API.
Happy to start with a paid diagnostic using data you already have.
Comment · 4 months ago
Interesting work. Is this something where SAM-3D or it background applies to?
Comment · 8 months ago
This is great and exciting. I happened to be doing some research to build memory-efficient diffusion models. I have not yet built the demo, but looking at a mix of architecture from several papers, IMTalker, SageAttension, FlashVSR, and Sparse VideoGen, with the intention to reduce memory to about 8GB.
The plan was to swap FlashAttention out, and also for an audio driver; SVG could have improved. At 60FPS, I think you are already doing this.
Great work.
Comment · 9 months ago
I have been working on CensorX, a multimodal content moderation set of models. It is from a personal project where I built content moderation in a Discord Bot.
I have open-sourced the fine-tuned models on Hugging Face and am looking for feedback on false positives/negatives in real-world scenarios.
The main exploration has been ablations based on freezing certain layers of the transformers. More work could be explored by tuning other parameters and expanding the datasets.
The Models: • Image (ViT-B/16): Fine-tuned Vision Transformer achieving 91.9% Accuracy and 0.99 AUC. o Link: https://huggingface.co/acaciabengo/nsfw_image_detection • Text (DistilBERT): Binary classifier trained on ~200k samples. o Focus: Optimized for low-latency inference (<100ms) to fit into real-time chat streams. o Link: https://huggingface.co/acaciabengo/nsfw_text_detection How to try it: 1. Self-Host (Free): You can pull the weights directly from Hugging Face and run them in your own container. 2. Managed API (Freemium): I have deployed these exact models as a high-availability API on RapidAPI. There is a free tier for testing. RapidAPI I am very interested in feedback on: • Performance • Access to larger datasets • Shared experience from people who have handled similar tasks Thank You
https://www.loom.com/share/b741bc127c814db395832b01bd086e96