Founder · Engineer · Educator

Aansh Shah

Founder and engineer building AI infrastructure for high-stakes work. I lead Briefcase AI, serve as President of Speech Oak Inc, and previously built security, observability, privacy-preserving machine-learning, and petabyte-scale data systems at Amazon Alexa and AGI. I hold an MS in Computer Science and a BS in Applied Mathematics and Computer Science from Brown University.

LinkedIn GitHub

Current work

Companies

Founder & CEO, Briefcase AI

Visit Briefcase AI

Briefcase AI helps organizations use artificial intelligence safely by enforcing their rules, recording how decisions were made, and producing evidence that teams, auditors, and regulators can verify.

Governance infrastructure

An open-source Python, Rust, and TypeScript SDK for controlling, recording, replaying, and auditing AI decisions.

Regulatory operations

A source-backed compliance lakehouse, human-confirmed Notebook CoPilot, and bank modules for BSA/AML, Metro 2 and Reg V, Reg E, fair lending, Reg Z, and BaaS change management.

Legal AI evaluation

Deterministic contract-agent benchmarks and verifiable legal-reasoning environments for evaluation and post-training.

President, Speech Oak Inc

Visit Speech Oak

Speech Oak Inc builds software that helps speech-language pathologists run engaging therapy sessions while automatically turning session activity into notes, progress records, and parent updates.

Speech Oak Planner combines game-driven articulation practice with clinician planning tools and child-friendly experiences. It captures trial data as therapy happens, then prepares SOAP notes, IEP progress reporting, and parent-ready updates so clinicians can spend less time recreating documentation after a session.

Selected experience

Experience

Software Development Engineer

Amazon Alexa and Artificial General Intelligence

  • Designed an identity and access platform processing 8+ million authentication requests daily at 99.999% availability. Automated workflows cut provisioning from 5 days to under 3 hours, and a claims-based portal enabled granular, federated access to 55+ critical datasets.
  • Audited 15 services handling PII and sensitive data with zero high-severity penetration-test findings, delivered 30 security consultations across 20 teams, and created reusable frameworks that cut review cycles by 45% and security incidents by 28%.
  • Built a privacy-preserving continual-learning platform for voice-recognition models, reducing deployment time by 30%, enforcing a 24-hour data-deletion policy, and replacing 24+ hour reporting with 2-minute performance metrics.
  • Built an OpenTelemetry observability platform that reduced customer-feedback processing from 8 hours to 3 minutes. Its zero-code Python SDK cut integration effort by 92%, reduced escalations by 15%, saved retail teams 15+ hours each week, expanded prompt-testing capacity by 30%, and supported $3.2 million in quarterly revenue.
  • Redesigned petabyte-scale data processing with a Rust compaction service that cut S3 throttling by 90%, infrastructure costs by 63% ($1.2 million annually), and incident recovery from hours to minutes, while nearly doubling throughput and improving customer satisfaction from 3.0 to 4.5.

AI Education Lead and Curriculum Developer

Inspirit AI

Teach artificial intelligence through applied projects in healthcare, climate, and social impact; lead individual research mentorship; and develop curriculum that makes machine learning, computer vision, natural-language processing, and responsible AI accessible to high-school students.

Research and teaching

Research and Teaching

AMIA Clinical Informatics Conference · 2020

Comparing Global with Disease-Specific Machine-Learned Readmission Prediction Models

First-author research on ICU readmission prediction using clinical data and machine learning.

Read the paper

NeurIPS KR2ML Workshop · 2020

Graph Embedding Priors for Multi-Task Deep Reinforcement Learning

Co-authored research on using knowledge-graph embeddings as linguistic priors for reinforcement-learning agents.

Read the paper

International Conference on Social Robotics · 2019

Stackelberg Punishment and Bully-Proofing Autonomous Vehicles

Research on game-theoretic policies that help autonomous vehicles negotiate safely with human drivers.

Read on arXiv

Conference talk · 2023

Scaling Real-Time Stream Processing for Machine-Learning Pipelines

Presented at Apache Flink Forward on operating real-time data systems for machine-learning workloads.

Watch the talk

Education and service

Education and Impact

Brown University · 2020

Master of Science

Computer Science

Brown University · 2019

Bachelor of Science

Applied Mathematics and Computer Science

Community service · 2011 to Present

Founder, GameCraft Charity

Organized more than 25 community events and distributed more than 500 gaming systems to help underprivileged young people share in the social and creative experiences of play.