Daniel Manning

Software Engineer,

Co-founder.

Daniel Manning — Software Engineer

Hey - I'm Dan.

I design and build full-stack applications, focusing on seamless user experiences and impactful, high-quality projects.

Experience

Atlas (Aug 2025 - Current)
Co-founder

What started as "an LLM with course content for students"  has now gained traction at RMIT and gotten validation in a ~350 student experiment. My business partner, Ed, and I are now successful candidates in RMIT's LaunchHub program, and are doing everything we can to find safe ways to use AI in education that preserve, and accelerate, learning quality and academic integrity.

Logo

Primary (Aug 2024 - Jun 2025)
Senior/Lead Software Engineer

RemixReactTypeScriptTailwindNodeJSPostgreSQLPrisma
  • Spearheaded front-end development using React + TypeScript in Remix, collaborating with designers to build polished, user-centric interfaces.
  • Delivered impactful features across the full stack by bridging technical gaps, proposing innovative solutions, and maintaining a “get-it-done” mindset.
  • Led and planned product-delivering projects in collaboration with engineering and design.
Logo

ReadCloud (Mar 2023 - Aug 2024)
Senior/Lead Software Engineer

ReactTypeScriptTailwindNodeJSDenoAWSPostgreSQLMongoDBPHPJavaKotlin
  • Led a small team of engineers through planning, mentoring, and hands-on development of core projects, including a full-stack LMS plugin used by ~10k active users.
  • Initiated AWS cost savings and optimizations, and the migration of existing infrastruction to CDK + SST.
  • Introduced observability infrastructure like CloudWatch, Sentry, and analytics using PostHog.
  • Maintained and/or migrated legacy PHP, Python, and Java + Kotlin systems.
  • Exposed to a wide variety of tools and frameworks, including Tauri, React Native and Expo as native app solutions. Additionally, education management infrastructure such as Canvas, Moodle, SchoolBox, and VETtrak in particular.
Logo

CultureAmp (Nov 2019 - Mar 2023)
Senior Software Engineer

ReactTypeScriptNodeJSAWSKotlinElm
  • Contributed to full-stack development projects involving React, TypeScript, NodeJS, and Kotlin.
  • Maintained the in-house design system (Kaizen Design System) and optimized development processes across teams by automating dependency updates using Renovate.
  • Collaborated on SRE-focused projects involving AWS and CDK, enhancing system reliability and deployment efficiency.
  • Worked on event-sourcing architectures and CQRS-based applications.
Logo

ReadCloud (Oct 2017 - Nov 2019)
Software Engineer

ReactTypeScriptNodeJSMongoDBAWSJavaKotlin
  • Developed scalable features using React, TypeScript, NodeJS, and MongoDB.
  • Migrated infrastructure from Docker-based deployments on vanilla VMs to AWS-backed automated CI/CD pipelines.
  • Designed intuitive user interfaces in collaboration with designers, improving product adoption.
Logo

Airwallex (Nov 2016 - Oct 2017)
Junior-Mid Software Engineer

ReactTypeScriptFlowDockerJavaKotlinCassandraDB
  • Mostly frontend engineer with exposure to Kotlin, Java + Spring, React, Typescript + Flow.
  • Gained experience with distributed systems such as Cassandra DB, RabbitMQ
  • Became adept with Docker (incl. Compose and Swarm, and some Kubernetes)

Education

Logo

Bachelor of Science (Computer Science) Honours 1st Class
Jan 2020 - June 2021
RMIT

Computer science honours degree with a focus on research, completing a 6 month thesis on the topic of fairness in multi armed bandits (in the recommendation domain), along with engaging electives involving AI techniques in games (heavy focus on reinforcement learning), and cloud infrastructure development.

Transcript | Thesis

Course
Mark
Honours Research Project (Thesis)
89%
Research Methods
86%
Games and AI Techniques
95%
Cloud Computing (AWS and GCP)
95%

GPA: 4.0


Later in 2022, I began a PhD into the intersection between deep reinforcement learning (e.g. PPO) and large pretrained language models (e.g. GPT, BERT), to understand how we can improve overall reasoining capabilities in sequential decision making. I intensively study how these language models perform and can be used to play text-based games like Zork, and, at the time, found unanswered research questions relating to their inability to determine which actions they can make, and what their long term effects are. My passion lies here: to give language models the ability to think slowly rather than quickly.

Logo

Bachelor of Software Engineering
Jan 2013 - Nov 2016
RMIT

TranscriptGPA: 4.0

Projects

Logo

Umi Chatbot (Aug 2025 - Current)
Engineer and Research Assistant

This project is about uplifting Umi - a chatbot for helping people navigate online abuse of personal and sensitive images. It's packed with useful knowledge about what to say and when, and about relevant legislation. We're now improving it's user experience, ability to collect more meaningful and researchable data, and use lessons learned making Atlas to greatly improve information retrieval and conversational quality.

Logo

The Plant Broker (Jan 2025 - Current)
Consulting Engineer

A Shopify store selling plants online to metro and regional Melbourne, where I develop backend automation solutions to solve stock/logistical problems that are key to the business' success. This project gave me a lot of experience in scraping, background jobs in NextJS, and Shopify's domain models (their GraphQL schema in particular).

Equalised Odds is not Equal Individual Odds: Post-processing for Group and Individual Fairness (FAccT 2024)
Co-author

This paper highlights a fundamental tension between group fairness (e.g. accuracy between demographics) and individual fairness (e.g. accuracy between any two people). We show that widely used existing post-processing methods involving randomisation violate individual fairness and prevent some groups from ever accessing the same classification odds as others. To address this, we introduce a new fairness notion called equalised individual odds and propose continuous, monotonic probability functions constrained by Lipschitz continuity to replace step-like decision rules. This method ensures smooth treatment across thresholds, preserves predictive accuracy, and guarantees that all individuals can access the full range of classification odds available to any group. The approach demonstrates, on widely used test sets, that it can maintain group fairness while improving individual fairness and transparency, thereby reconciling the conflict between these two fairness objectives.

Referees available on request