J
Joseph Lord
About
Detail
California, United States
Résumé
Jobs
verified_user
0% verified
- CSenior AI/ML EngineerCONTEXTUAL AIFeb 2024 - Current (2 years 8 months)Built a retrieval-augmented generation (RAG) system to help Qualcomm engineers quickly find answers in 3GPP standards, patents, and internal documents. • Ingested and indexed millions of documents using FAISS-GPU, paired with BM25 in OpenSearch. • Fine-tuned large language models on telecom terminology with Hugging Face Transformers and evaluated accuracy with domain-specific benchmarks. • Deployed models on Kubernetes with NVIDIA Triton/TensorRT for fast, low-latency inference. • Monitored performance and hallucination rates with Prometheus and Grafana dashboards and real-time logging. • Integrated the assistant into Qualcomm's engineering portal, returning sources with every answer. • Reduced research time from hours to seconds with
Senior Machine Learning EngineerWestern Governors UniversityApr 2022 - Feb 2024 (1 year 11 months)CoachDesk - Multi-Agent AI System Prototype • Set up AI agents for different roles (tutor, grader, coach) using GPT-4 APIs, LangChain, and Python, enhancing the system's ability to automate educational tasks. • Created mock student data and built small databases with pandas and SQLite to test and improve AI's ability to personalize responses, leading to more tailored educational interactions. • Wrote prompts and tested conversation flows in Jupyter notebooks to ensure seamless collaboration among agents, improving the overall user experience. • Built simple connections between agents with FastAPI and Redis, enabling indirect communication and enhancing system integration. • Packaged prototypes with Docker and used GitHub Actions and AW- Machine Learning EngineerKasistoSep 2020 - Mar 2022 (1 year 7 months)AI-Powered Hackathon Assistant • built retrieval-augmented chatbot using embeddings and GPT-3 to provide real-time suggestions for project descriptions, titles, and tags. • Implemented FastAPI to connect the chatbot with a retrieval system for fetching hackathonspecific FAQs and guidelines, enhancing user query relevance. • Integrated sentence embeddings for similarity searches to retrieve the most relevant content, ensuring responses were specific and contextually accurate. • Leveraged Hugging Face Transformers to fine-tune LLMs for generating hackathon announcements, FAQs, and project summaries. • Used MongoDB to store user profiles, chatbot interaction logs, and submission data, enabling real-time retrieval and personalized assistan
- NMachine Learning EngineerNarrativeDxJul 2018 - Aug 2020 (2 years 2 months)• Designed and deployed AI-driven chatbots using Microsoft Bot Framework to automate customer interactions and support workflows. • Built NLP pipelines for intent detection, entity recognition, and sentiment analysis using spaCy, NLTK. • Enhanced language understanding capabilities with word embeddings (Word2Vec) and pre-trained transformer model BERT, adapting them for domain-specific tasks. • Added multilingual support using Google Translate API and custom pipelines for non-English text processing. • Built RESTful APIs using Django to handle requests from chatbots and manage communication with backend services. • Integrated Django with PostgreSQL using SQLAlchemy, improving data retrieval and storage efficiency for faster applicatio
- Computer Vision EngineerBairesDevJan 2015 - Jun 2018 (3 years 6 months)• Designed machine learning models for disease prediction and medical diagnostics using Pythscikit-learn, and early versions of TensorFlow and Keras, improving diagnostic accuracy in pilot testson. • Built custom CNNs for medical image classification and enhanced performance with transfer learning using AlexNet and VGG16, resulting in faster and more accurate image analysis. • Developed NLP pipelines with spaCy, Word2Vec, and NLTK to analyze clinical notes and extract structured medical entities like symptoms and medications, streamlining data processing for clinical research. • Implemented early object detection systems using Faster R-CNN and custom Python scripts to identify anomalies in medical imaging, enhancing early detection capa
- PMachine Learning InternPinewood AnalyticsJan 2014 - Dec 2014 (1 year)• Developed image classification and object detection models using MATLAB, ImageJ, and CNN frameworks. • Preprocessed and annotated medical imaging datasets with Python tools and optimized model performance using PCA and hyperparameter tuning. • Built and deployed a prototype system that improved biomarker detection accuracy, reducing false negatives in medical imaging.
- FFinancial Knowledge AssistantFinScopeDeveloped an AI assistant that retrieves and summarizes financial research, market news, and policy docs for analysts and compliance teams. • Built ingestion pipelines for market research, financial news, and compliance manuals using PySpark and Databricks, enhancing data processing efficiency and accuracy. • Indexed all financial documents in Pinecone's private cloud, added filters like date and author, and used both keyword search and AI embeddings together to make results more accurate and trustworthy. • Fine-tuned financial domain language models with Hugging Face and MLflow, improving the safety and professionalism of automated communications. • Added compliance guardrails including audit logging, role-based access control, and out
Education
verified_user
0% verified
- AAWS Certified Cloud Practitioner CISCO Certified CyberOps Associate
- BB.S. in Computer Science
- M.S. in Computer ScienceRice University