Siftain
Ahmad
I build ML and LLM systems that run in production — RAG pipelines, agentic workflows, and FastAPI services shipping to real users, not demos.
"Machine Learning Engineer with 2 years of experience shipping ML, LLM, RAG and agentic AI systems, skilled in vector search and scalable ML APIs using Python, FastAPI, Docker and AWS."
Experience
- Designed, built and deployed production LLM and ML systems end to end, cutting inference cost by 67% and latency by 36% across live products serving 50+ restaurants.
- Processed 10,000+ customer support tickets into a 5-category NLP routing pipeline.
- Fine-tuned a DistilBERT classifier reaching 89% Macro F1, beating TF-IDF + Logistic Regression/SVM baselines.
Featured Projects
Multimodal menu-intelligence and recommendation platform. Extracts structured data from menu images, then serves multi-turn dish recommendations via hybrid retrieval. Serves 50+ restaurants in production.
Real estate price-prediction platform on 1M+ property records with geospatial and temporal feature engineering. Random Forest was the best performer, deployed via FastAPI.
Agentic lead-generation and outreach pipeline combining 3 LLM providers, live web search, and 16+ RSS feeds, with scoring, dedup, and enrichment before outreach.
Research
17-layer CNN trained on nearly 80,000 MRI images spanning four stages of Alzheimer's disease progression, using preprocessing and data augmentation to improve generalization — reaching ~99% accuracy with high precision and recall.
Certifications
Completed program certification
