Hello, I am █
Engineering the future of AI/ML — where clinical precision meets computational intelligence.
Kinetrix Technologies LLP
Built clinical speech-to-text (Whisper, BERT) and radiology image analysis (YOLO, ResNet) modules for CARE, an open-source healthcare platform, served through a FastAPI backend secured with OAuth2, JWT and RBAC, with inference on Celery workers.
Infotact Solutions (Virtual)
Ran end-to-end ML workflows on real-world datasets (cleaning, feature engineering, training, evaluation) with pandas and scikit-learn, comparing models on held-out metrics.
SCANNING REPOSITORIES...
01_FEATURED_STABLE
Private, local document Q&A with citations (FastAPI, FAISS, Ollama). Retrieval tested with recall@k and MRR.
02_STABLE
Turns synthetic doctor notes into structured sections and a medication list. F1 0.71 → 0.97 on held-out notes.
03_STABLE
Flood damage maps from drone images with a U-Net + ResNet34 in PyTorch. 70.7% mean IoU on 448 test images.
04_STABLE
Cafe recommender using TF-IDF similarity and ratings, redesigned after a data audit ruled out collaborative filtering.
01_FEATURED_STABLE
Private, local document Q&A with citations (FastAPI, FAISS, Ollama). Retrieval tested with recall@k and MRR.
02_STABLE
Turns synthetic doctor notes into structured sections and a medication list. F1 0.71 → 0.97 on held-out notes.
03_STABLE
Flood damage maps from drone images with a U-Net + ResNet34 in PyTorch. 70.7% mean IoU on 448 test images.
04_STABLE
Cafe recommender using TF-IDF similarity and ratings, redesigned after a data audit ruled out collaborative filtering.
Python
SQL
PyTorch
Scikit-learn
Hugging Face
OpenCV
RAG
FAISS
Ollama
Embeddings
FastAPI
Docker
Celery
Git