Aditya Singh

Aditya
Singh

CS @ UCSB

SWE Lead @ Ryft AI

Building Ziri, a voice OS

Robotics backgroundCambridge researchIEEE MIT URTC
ProjectsResume

Voice

Ask by voice.

Live ElevenLabs agent trained on my work, projects, and background.

Projects
Ryft
Ziri
Email instead

Requires mic access. Chrome or Safari work best.

Live voice

waiting

Start talking.

Click the button, allow your mic, and ask a real question.

What do you like building most?
Which project is your favorite?
What kind of work do you want next?
What matters to you when you build AI?

About

CS at UCSB. SWE Lead at Ryft AI.

Building Ziri, a voice OS. Background in robotics, computer vision, and ML research at Cambridge.

I mostly build full stack infra, voice products, and agent systems.

Ryft

Building commission automation infra and multi-agent backend systems.

Side projects

Ziri, workflow tools, and whatever AI product idea I am trying next.

Vibecoding

Basically stress-testing agentic IDEs and seeing how much real work I can get out of them.

School

UCSB CS, class of 2027

Ryft AI

work

sales compensation infra

Robotics

background

nonprofit + competition bots

Soft robotics

research

Cambridge + ML

IEEE MIT URTC

paper + talk

published + presented

Stack

Languages

Python
TypeScript
Rust
Go

Frameworks

FastAPI
Next.js
React
Node.js

Infra & tools

AWS
Docker
Postgres
Linux

Experience

Work, robotics, and research.

The main thread is building real systems and making them useful to other people.

Aug 2025 — Present

Cupertino, CA

Ryft AI

Software Engineering Intern

Building backend and agent systems for commission automation. A lot of the work is turning messy contracts into software people can trust.

  • Architected a scalable Python/FastAPI backend spanning 40 REST endpoints for a 3-tier multi-agent orchestrator.
  • Designed an autonomous tool-driven loop for extracting commission logic from messy PDF contracts with high-accuracy structured output.
  • Engineered pgvector-based context systems and memory compaction to cut recurring LLM cost by 60-70%.
  • Built validation tooling that removed 350-400 hours of manual QA per quarter.

Proof points

40 REST endpoints

30k ARR

350-400 hours saved / quarter

Sep 2018 — Jan 2024

Fremont, CA

Techno G.O.A.Ts

Co-Founder & Team Captain

Co-founded a robotics nonprofit. Built competition robots, led the team, and mentored students.

  • 3D-modeled 30+ competition robots and ran weekly build sprints across mechanical, software, and strategy work.
  • Built a TensorFlow/OpenCV vision pipeline that reduced autonomous task latency by 40%+ at 96% detection accuracy.
  • Mentored neurodivergent learners through a 4-year Serendipity STEM partnership and helped secure a $5k Apple grant.

Proof points

30+ robots modeled

4 years mentoring

$5k Apple grant

2023

Cambridge, UK

University of Cambridge / IEEE

Soft Robotics Research Collaborator

Worked on soft robotics research at Cambridge and turned it into published work.

  • Worked on learned forward-kinematics approaches for soft continuum robots where analytical models break down.
  • Formalized the project into research rigorous enough to publish and present outside the lab.
  • Used the research process itself as a training ground for making complex technical systems explainable.

Proof points

IEEE publication

MIT URTC presenter

Cambridge collaboration

Education

University of California, Santa Barbara

B.S. in Computer Science. GPA 3.97. Expected graduation 2027.

3.97

GPA

2027

Expected graduation

Relevant coursework

Data Structures & Algorithms
Computer Architecture
Object-Oriented Programming
Linear Algebra
Differential Equations
Multivariable Calculus
Problem Solving with C++

Outside class

Building products

Doing research

Shipping open source work

Projects

Pinned projects.

01
Voice OSUpdated Mar 11

Ziri

A distributed home voice OS that listens locally, routes hard tasks through agents, and keeps routine commands fast.

Creator2026Python
always-on listener
Mac control
smart-home actions
calendar + music tools
semantic memory
DetailsSource

sub-900ms

Complex queries

80%

Deterministic path

4

Input surfaces

PythonFastAPILangGraphAWS BedrockpgvectorLangfuse

Contact

Email works best.

If you want to build something together, reach out.

adityasingh@ucsb.edu

Location

Santa Barbara / Goleta

Looking for

product engineering, AI systems, voice products

Elsewhere

GitHubLinkedInEmail

Designed and built by Aditya Singh