Bryant Avila

Bryant Avila

AI Systems Engineer, ABT-PhD in physics, data & network scientist… also into surfing

Hi! 👋🏼 Please explore my website!

“In God we trust; all others must bring data.”
– W. Edwards Deming (American statistician).

Research Interests

  • Production AI systems, multi-agent orchestration, RAG, and LLMOps.
  • NLP, NLU, embedding spaces, and legal / enterprise search.
  • Machine learning, especially graph neural networks.
  • Reinforcement learning & integer linear programming optimization.
  • Brain, financial, social, & car-traffic networks.
  • Inferring network structure from its dynamics 🕸️ ↔ 〰️.

Recent Personal Projects

  • VisualSounds: Eye-Gaze Detection (private)
    • The first building stone of something new: an iOS system for on-screen eye-gaze detection (TrueDepth + ARKit). What follows is still under wraps — glimpses only — toward a new media experience powered by text-to-audio transformers.
  • Open-Forecast: A FinTech Tool
    • A fintech web app built with Streamlit, designed to forecast stock prices using machine learning. It analyzes pre-market data to predict price trends.
  • Contextual Search: A RAG Tool
    • A search engine powered by a fine-tuned DistilBERT model that enables contextual searches within movie scripts. This tool uses Python scripts for scene extraction and embedding generation.

Award

My team and I secured a prestigious National Science Foundation SBIR Phase I grant worth $275K in 2023 for pioneering an artificial intelligence system to enhance transparency and predict trends in democratic elections through sentiment analysis and network theories.

Work Experience

  • AI Systems Engineer - Full time, DBE Legal (Nov 2024 – Present) — DTLA, CA
    • Joined a team that knew they had a problem but couldn’t yet define a solution. Worked with the firm’s legal teams to understand their full workflow and key trigger points, then designed and shipped AI systems for searching massive case-file databases, tracking client matters, and monitoring defendant-level status. Managed multiple projects from start to finish.
    • Created a Svelte + Django interactive dashboard that is continually updated by LLM agents triggered when SQS detects file changes in the firm’s file systems.
    • Architected a production-grade AI legal research platform with multi-agent reasoning, a Milvus RAG system, GPU-accelerated model serving, and user-isolated SvelteKit frontend instances.
    • Built a comprehensive service mesh spanning 10+ FastAPI servers, including agent orchestration, backend coordination, GPU load balancing across 6 GPUs and cloud providers, task termination, SQL logging, and MCP tool servers.
    • Orchestrated a multi-agent architecture using a custom in-house graph-based framework (conceptually similar to LangGraph), where a planning LLM delegates tasks to specialized agents (paralegal answerer, cross-document summarizer, database inspector), executing parallel tool calls via an MCP server and reducing research time from hours to minutes.
  • Machine Learning Engineer - Contract, RivetAI (June 2024 – August 2024) — Culver City, CA
    • Spearheaded an industry-first dataset by leveraging AWS and Azure large language models (LLMs) to generate and augment data from movie scripts, enabling content extraction and analysis.
    • Employed advanced prompt engineering, minimizing paraphrasing by 7% and ensuring structured output in JSON format.
    • Fine-tuned a custom mini-BERT NER transformer model, increasing entity recognition accuracy by 15% and setting new benchmarks for script data processing.
  • Data Scientist - Contract, Kcore Analytics (Nov 2023 – May 2024) — Remote
    • Designed and deployed an ETL pipeline (Databricks) to track real-time electoral campaign performance, improving outcome predictions by 10%.
    • Preprocessed and aggregated 1TB+ of data using Spark and SQL, reducing data retrieval time by 20%.
    • Applied geospatial analysis and sentiment classification on tweets through PyTorch, predicting voting outcomes at the county level by training a graph neural network (GNN).
  • Physics Research Assistant - Full time, Levich Institute (Aug 2019 – Aug 2024) — New York, NY
    • Developed a GIS-based system to predict and manage rat infestations in Manhattan, improving prediction accuracy by 15% using NYC Open Data.
    • Built a MATLAB dashboard used by 30+ researchers for neuron network dynamic simulations, enhancing scalability and noise testing capabilities.
    • Optimized integer linear programming scripts through A/B testing based on graph theory with NetworkX and GUROBI in Python, achieving network repair with 10Ă— fewer modifications and reduced memory usage… (more in CV)
  • Laboratory Manager - Full time, Fashion Institute of Technology (Jan 2018 – Jan 2024) — New York, NY
    • Led a team in setting up research labs focused on fabrics research, ensuring timely, efficient execution of experiments and demonstrations.
    • Managed a $100K annual budget, securing funds to integrate 3D printing technology, enhancing student learning, and increasing equipment upgrades.
    • Spearheaded the development of sustainable materials, collaborating with faculty on bacterial leather and mycelium projects, reducing growth times by 5 days and delivering results on time for semester reports… (more experiences in my CV)
  • Molecular Data Scientist - Full time, Biochemistry Lab (April 2016 – Nov 2017) — New York, NY
    • Advanced virus morphology research by analyzing terabyte-scale biochemical datasets using Relion and R, producing atom-resolution 3D models that contributed to the development of potential treatments.
    • Innovated multi-processing techniques to integrate Protein Data Bank models with cryo-EM data, enhancing the accuracy of molecular structures and supporting breakthroughs in structural biology.

Education

  • ABT-PhD in Physics - The City College of New York - Feb 2019 – Dec 2024
  • M.S. in Physics - CUNY Graduate Center - Sept 2019 – Feb 2022
  • B.S. in Physics - The City College of New York - Feb 2012 – Feb 2016

Skills

  • Coding
    • AI / LLM systems: multi-agent orchestration, RAG (Milvus), MCP tools, prompt engineering, Hugging Face, FastAPI, Svelte / SvelteKit, Django
    • ML / data: Python (NumPy, SciPy, Pandas, scikit-learn, Matplotlib, TensorFlow, PyTorch, GeoPandas, NetworkX, Selenium, GUROBI), Spark, SQL, Databricks, R, MATLAB
    • Other: C++, Fortran, LaTeX, LabVIEW, Relion, Fusion 360, SolidWorks, GitHub, ArcGIS, AutoCAD, WordPress, Salesforce
  • Soft skills
    • Team management, cross-functional collaboration, public speaking, project management, problem-solving, attention to detail, quick learner

Publications