{ "$schema": "https://raw.githubusercontent.com/jsonresume/resume-schema/v1.0.0/schema.json", "basics": { "name": "Christopher Mejo, Ph.D.", "label": "AI Research Scientist — LLM & Reinforcement Learning Systems", "image": "", "email": "work@cmejo.com", "url": "https://cmejo.com", "summary": "AI research scientist and engineer with 15+ years taking machine learning systems from research prototype to production, holding a Ph.D. in Physics from Harvard (quantum information and complexity theory, completed at 21) and postdoctoral training at MIT under Scott Aaronson. Core focus: LLM systems — retrieval-augmented generation, fine-tuning, agentic reasoning — and reinforcement learning for sequential decision-making. Solo-architected AI Scholar, an open-source RAG platform indexing 15M+ research papers (Google Summer of Code Fellow, 2024 and 2025). Shipped production LLM and RL systems in finance and a federated computer-vision system in healthcare. Comfortable owning systems end-to-end in small teams.", "location": { "city": "New York", "region": "NY", "countryCode": "US" }, "profiles": [ { "network": "LinkedIn", "username": "cmejo", "url": "https://linkedin.com/in/cmejo" }, { "network": "GitHub", "username": "cmejo", "url": "https://github.com/cmejo" } ] }, "work": [ { "name": "Braverock LLC", "position": "Senior Data Scientist", "startDate": "2021-11", "endDate": "Present", "summary": "Quantitative ML research and production RL trading system architecture for high-frequency and multi-horizon market strategies.", "highlights": [ "Architected a reinforcement-learning trading system (deep Q-networks, actor-critic methods) for multi-asset portfolio optimization, improving Sharpe ratio 23% over prior strategy on microsecond-latency data, validated across both backtested and live trading.", "Built transformer-based market-prediction models for financial time series, deployed in a pipeline processing 1B+ quote-data points/day at microsecond-scale latency.", "Built a RAG system over 10M+ financial documents (SEC filings, earnings calls, research reports) for signal extraction, using custom embeddings and a production vector database.", "Directed distributed pretraining and instruction fine-tuning pipelines for domain-specific 7B–14B parameter foundation models (DeepSpeed ZeRO-3, custom CUDA memory-mapped tokenizers), reducing per-step wall-clock latency 38%.", "Implemented dynamic Heterogeneous Graph Attention Networks (HAN) modeling multi-asset cross-correlations across tick, intraday, and multi-week horizons, improving regime-shift detection accuracy 19%.", "Designed and deployed an enterprise crypto basis-yield strategy delivering an out-of-sample, beta-neutral Sharpe ratio of 2.63 on $60M+ in institutional AUM.", "Developed proprietary quantitative strategies achieving a 6.2 Sharpe ratio (Sortino 8.4, max drawdown <1.9%) over a continuous 24-month live deployment window, executing billions of dollars in trading volume across holding periods from sub-second to multi-month.", "Strategy scope spanned quantitative multi-exchange, multi-asset market making; treasury management; regime-aware risk-managed momentum; and multi- and single-asset statistical arbitrage / structural mean-reversion market-neutral strategies." ] }, { "name": "Sandia National Laboratories / Joint Google Quantum AI Initiative", "position": "Visiting Research Scientist (part-time)", "startDate": "2020", "endDate": "2024", "summary": "Quantum Computing & High Energy Physics research in a joint initiative with Google Quantum AI.", "highlights": [ "Senior researcher developing variational quantum eigensolvers for lattice gauge theory calculations.", "Designed quantum error correction protocols (surface codes) and resource-estimation frameworks for near-term quantum devices (Qiskit, Cirq).", "Ran distributed quantum many-body simulations (50+ qubits) on SLURM-managed clusters; contributed to DARPA and IARPA research proposals.", "Active Top Secret/Sensitive Compartmented Information (TS/SCI) security clearance." ] }, { "name": "R Foundation for Statistical Computing", "position": "Google Summer of Code Fellow", "startDate": "2024", "endDate": "2025", "summary": "AI Scholar Project (github.com/cmejo/AI_Scholar), an open-source hybrid RAG platform indexing 15M+ research papers.", "highlights": [ "Architected AI Scholar, combining dense (FAISS), sparse (BM25), and graph-based (Neo4j) retrieval for scientific literature search across 15M+ research papers.", "Built an adaptive retrieval system using meta-learning to personalize search strategy, improving relevance 40% (NDCG@10) over baseline on a 200-query held-out set with citation-derived relevance judgments.", "Implemented chain-of-thought reasoning over the RAG pipeline with explicit source attribution and calibrated uncertainty (Monte Carlo dropout, evidential deep learning).", "Built a dynamic knowledge-graph construction pipeline using named entity recognition (spaCy, BERT-NER) and relation extraction to connect concepts, authors, and methodologies." ] }, { "name": "Cushion AI", "position": "Senior Data Scientist", "startDate": "2021-04", "endDate": "2021-10", "summary": "Transformer NLP and transaction categorization for consumer fintech.", "highlights": [ "Built CushionNLP, a proprietary BERT-based transformer system for bank-fee negotiation, driving a 30% increase in successful negotiations and $2M+ in additional annual customer savings.", "Architected a transaction-categorization system (ensemble methods, graph neural networks) achieving 95% accuracy across 80M+ transactions from ~300 financial institutions, cutting processing time 50% versus the prior rule-based (non-ML) system." ] }, { "name": "RapidRads AI", "position": "Research Data Scientist", "startDate": "2019", "endDate": "2021-02", "summary": "Medical computer vision and federated learning under differential privacy.", "highlights": [ "Built a computer-vision system (EfficientNet-B7 backbone) for COVID-19 detection from chest X-rays, achieving 0.972 ROC-AUC with 96.5% sensitivity and 98.2% specificity (1.8% false-positive rate, 3.5% false-negative rate) under differential privacy (ε = 1.2, δ = 10^-5) across a federated-learning deployment spanning 4 medical institutions.", "Combined multiple CNN architectures via ensemble methods with Grad-CAM visualization and uncertainty quantification for clinical validation." ] }, { "name": "Confiant", "position": "Senior Data Scientist & Cybersecurity Researcher", "startDate": "2017", "endDate": "2019", "summary": "Large-scale ad tech cybersecurity and graph anomaly detection.", "highlights": [ "Built threat-detection models (Isolation Forest, One-Class SVM, LSTM autoencoders) and GraphSAGE-based botnet-detection models identifying coordinated attacks across 1B+ daily ad requests." ] }, { "name": "Cronode", "position": "CEO & Lead Architect", "startDate": "2011", "endDate": "2017", "summary": "Information retrieval and recommendation architectures.", "highlights": [ "Built a recommendation engine (TF-IDF, cosine similarity, collaborative filtering) and custom Solr ranking configuration." ] } ], "education": [ { "institution": "Massachusetts Institute of Technology (MIT)", "studyType": "Postdoctoral Fellowship", "area": "Theoretical Computer Science", "startDate": "2010", "endDate": "2012", "gpa": "", "courses": [ "Faculty Advisor: Dr. Scott Aaronson" ] }, { "institution": "Harvard University", "studyType": "Ph.D.", "area": "Physics (Quantum Information and Complexity Theory)", "startDate": "2007", "endDate": "2010", "courses": [ "Degree completed at age 21" ] }, { "institution": "Harvard University", "studyType": "B.S.", "area": "Physics and Computer Science", "startDate": "2004", "endDate": "2007" } ], "skills": [ { "name": "Programming Languages", "level": "Expert", "keywords": ["Python", "C", "C++", "R", "Assembly", "Java", "JavaScript"] }, { "name": "Machine Learning & Deep Learning", "level": "Expert", "keywords": ["PyTorch", "TensorFlow", "JAX", "Transformers", "Reinforcement Learning (DQN, actor-critic, MAML)"] }, { "name": "LLM & RAG Systems", "level": "Expert", "keywords": ["Retrieval-augmented generation", "FAISS", "BM25", "Neo4j knowledge graphs", "custom embedding models", "vector databases", "LangChain", "LangGraph", "vLLM", "spaCy", "BERT-NER", "chain-of-thought and agentic reasoning", "uncertainty quantification", "retrieval evaluation (NDCG, MRR)"] }, { "name": "MLOps & Distributed Systems", "level": "Expert", "keywords": ["MLflow", "Kubeflow", "Apache Airflow", "Docker", "Kubernetes", "Ray", "Weights & Biases", "SLURM", "multi-GPU distributed training"] }, { "name": "Quantum Computing", "level": "Advanced", "keywords": ["Qiskit", "Cirq", "Surface codes", "Variational Quantum Eigensolvers"] } ], "awards": [ { "title": "Google Summer of Code Fellow", "date": "2024", "awarder": "Google / R Foundation for Statistical Computing", "summary": "Selected for AI Scholar Project" }, { "title": "Google Summer of Code Fellow", "date": "2025", "awarder": "Google / R Foundation for Statistical Computing", "summary": "Selected for AI Scholar Project" } ], "meta": { "clearance": "Active Top Secret/Sensitive Compartmented Information (TS/SCI)", "version": "1.0.0", "lastModified": "2026-10-08" } }