jakpakoun.com

Software and Machine Learning Engineer

Jean-Jacques Akpakoun

M.S. in Artificial Intelligence student at the University of Nebraska at Omaha and production-trained software engineer. I build the full system around a model: simulation, data, training, evaluation, serving, and the platform that carries it.

Download resumePlay against the model

Experience

Work

Mar 2026 to present

AI Agent Systems Developer

University of Nebraska at Omaha, AI-CCORE

  • Contributed to a configurable no-code AI-agent platform for designing and deploying structured workflows.
  • Built the infrastructure around agent-generated applications: persistence, artifacts, deployment, workflow recovery and validation.
  • Deployed containerized AI-platform services behind a TLS reverse proxy with health checks and persistent runtime storage.

May 2024 to May 2025

Full-Stack Software Engineer

Gallup, Inc.

  • Designed and implemented a two-tier Redis caching system for client, survey and configuration data, reducing average page-load time by 65% (2.0s to 0.7s).
  • Eliminated backend query timeouts and cut execution time about 6x (60+ seconds to about 10) through SQL tuning, indexing and event batching in a .NET service.
  • Supported a 750K+ participant survey workload under peak load.

Jun 2022 to May 2024

Software Developer

American National Bank

  • Built a role-aware real-estate valuation platform with image workflows and aggregate analytics, reducing appraisal time-to-approval by 60%.
  • Developed secure ETL pipelines feeding loan and approval data into Power BI, with access controls and audit-oriented logging.

Machine learning

Platform and Research

2025 to present

Partially observable decision-agent and ML experimentation platform

Built on Pokemon Showdown. The trained policy is live at battle.jakpakoun.com.

  • Extended Pokemon Showdown into a controlled multi-agent environment for partially observable sequential decision modeling.
  • Built deterministic pipelines from raw battle logs through state reconstruction, dataset generation, model training, checkpointing and held-out evaluation, over approximately 200K battle turns.
  • Designed structured representations for entities, state, history, hidden information and legal actions.
  • Implemented Keras neural policies with multi-head action prediction and legal-action masking.
  • Served trained policies through live inference for interactive human evaluation.
  • Developed automated agent-vs-agent benchmarking and behavioral comparison infrastructure.

In progress

Iterative computation inside learned representations

  • Investigating whether repeated learned transformations of latent state can improve downstream prediction before an external decision is produced.
  • Designed frozen-component experiments isolating learned latent transformations from encoder and readout behavior.
  • Compared repeated latent computation against simpler corrective baselines to isolate the source of observed improvements.
  • Investigated memory-constrained sequence processing across dynamic programming, recurrent networks, attention and state-space models.

Teaching and delivery

Next-Gen Studio and Teaching

Summer 2026

Next-Gen Studio

  • Built and operated student-facing AI development infrastructure supporting multiple teams from early concept through deployed application.
  • Mentored teams on feasibility, architecture, scope, debugging and staged AI application delivery.
  • Led technical debugging, workflow recovery and application strengthening before the public showcase.
  • Delivered a public-facing showcase presenting nine Summer 2026 student applications.

Ongoing

Teaching

  • Tutor undergraduate students in data structures, algorithms, Python, JavaScript and C#, with an emphasis on debugging and computational problem solving.
  • Translate advanced technical concepts into progressive, project-driven learning experiences.
  • Adjunct faculty tutor at Metropolitan Community College (Nov 2020 to May 2021): Pre-Calculus and Statistics with personalized lesson plans.

University of Nebraska at Omaha

Education and Skills

M.S. in Artificial Intelligence

Computer Vision concentration. In progress, expected December 2027. GPA 4.000.

  • Completed: Design and Analysis of Algorithms, Fundamentals of Deep Learning, Ontologies in NLP and AI.
  • Fall 2026: Advanced Topics in Artificial Intelligence, Pattern Recognition.

B.S. in Computer Science

Artificial Intelligence concentration. May 2024. GPA 3.907, Summa Cum Laude. Dean's List every semester.

Skills

  • Programming: Python, SQL, C#, JavaScript, TypeScript, HTML, CSS
  • Machine learning: Keras, TensorFlow, NumPy, Pandas, scikit-learn
  • Software: ASP.NET, React, Node.js
  • Data and cloud: AWS (EC2, Lambda, SQS, SNS, Step Functions), Docker, Kubernetes, Redis, PostgreSQL, MySQL, DynamoDB, Power BI