Aaron SchlosserContact

Aaron John Schlosser, PhD

Engineering leader.
Literary scholar.

I’ve led software work in healthcare and election systems. I built DerridAI to help scholars trace AI-generated answers to the passages that support them.

View DerridAI case study Engineering experience

San Francisco Bay Area

FIG. 01DERRIDAI / EVIDENCE TRACE
How DerridAI links a claim to its source
01 Featured project

Independent academic project · Open source

DerridAI

A local-first workspace for AI-assisted research. Find relevant passages, ask questions across a corpus, and check each generated claim against the source.

Explore DerridAI View source on GitHub
My role
Creator and developer
Application stack
Python, FastAPI, Vue 3, TypeScript
Data and runtime
SQLite, ChromaDB, Docker Compose

Research workflow

From a research question to evidence you can cite

Find relevant passages
Search by keyword or meaning across works and languages. Narrow the results to the texts you want to examine.
Compare readings
Choose the passages the model works from, then ask where the texts agree, differ, or qualify a position.
Verify before citing
Open the supporting excerpt to check its wording, location, and attribution. This helps catch fabricated claims and citations that point to the right page but credit the wrong person.

I started with a multilingual corpus of Derrida’s work, where a single passage might quote, reinterpret, or dispute someone else’s argument. You can use the same workflow with your own corpus.

Fictional example

Trace a claim

Follow a generated sentence back to its source.

Claim

01 / Generated answer

“The essayist questions the historian’s claim that exile has a single cause.”

The answer describes both a position and someone else’s response to it.

Attribution

02 / Roles in the passage

Speaker
The essayist
Position holder
The historian
Stance
Questions the historian’s claim
Source

03 / Supporting passage

“The historian insists exile has one cause. The accounts that follow call that certainty into question.”

Sample essay, p. 18, ¶ 2

Research specification

cELF 1.0

Capta-Enriched Locatable Factum

I wrote cELF to give research tools a common way to record what evidence was used, where it came from, and which claim it supports. DerridAI is its reference implementation.

Read the cELF specification
Source
The document and version consulted.
Record
A passage with a source location and a revision history.
Evidence
The exact excerpt or span used in the research.
Claim
A statement with an explicit link to its supporting evidence.

How I built it

  1. 01 / Corpus Builder

    Review before publishing

    I split ingestion, segmentation, metadata enrichment, and review into distinct stages. Reviewers can inspect proposed changes against source passages and decide what becomes part of the corpus.

    Corpus review code
  2. 02 / Retrieval

    Keep the index rebuildable

    ChromaDB holds derived search data. The application keeps source documents and reviewed records separately, so rebuilding the index leaves the scholarly record intact. Hybrid search and reranking select passages for the model.

    Read the architecture
  3. 03 / Provenance

    Resolve citations in code

    The model synthesizes the selected passages. Application code resolves source locations, checks claim-to-evidence links, and renders citations.

    See the cELF model

Run it locally

Start the app with Docker Compose to build and query your own corpus. Fixed-case benchmarks let you compare pipeline configurations using the same inputs.

Setup instructions Benchmark tests
02 Engineering

Healthcare and election software

I’ve worked across patient-facing mobile platforms, healthcare analytics, voting systems, and event-driven infrastructure.

Engineering writing · 2016

How NATS powered real-time auctions

I wrote about how NextGen Leads used NATS to route events through normalization, bidding, and delivery in a Node.js microservices architecture.

Read the NATS article

  1. Healthcare technology

    Gozio

    Lead Software Engineer

    Patient-facing mobile software and hospital wayfinding.

    2021–2026

  2. Healthcare analytics

    Interpreta

    Senior Software Application Engineer

    2019–2021

  3. Election technology

    Votem

    Technical Lead, Process Architecture

    2018–2019

  4. Election technology

    Everyone Counts

    Technical Lead, Voting Systems

    2016–2018

  5. Web applications

    The Control Group

    Application Developer

    2014–2016
03 Scholarship

Literature, ethics, and exile

My training in French and comparative literature led me to study how Francophone writers represent exile and how they quote, reinterpret, and challenge other voices.

That kind of close reading became a software requirement in DerridAI. Readers need to see whose position an answer describes and open the passage that supports it.

Doctoral dissertation

Yale · 2016

The Ethics of Literature

The dissertation examines representations of exile in Abdelkébir Khatibi, Nabile Farès, Hélène Cixous, and Jacques Derrida. Christopher Miller and Edwige Tamalet Talbayev directed it.

Yale dissertation listing
Yale University
PhD in French, MPhil, and MA. I also taught in the French department.
École normale supérieure
Visiting research fellow, Paris.
Dartmouth College
AB in comparative literature, magna cum laude.
04 Contact

Let’s talk.

Find me on LinkedIn for conversations about engineering leadership and research software.

Connect on LinkedIn Browse my GitHub