Build the Future With Us
Join our team of student researchers and engineers pushing the boundaries of artificial intelligence.
Why Join TAI Research?
Real Impact
At TAI, you won't be working on theoretical projects. You'll be building and shipping real AI products used by developers around the world.
Learn by Doing
We believe the best way to learn AI is to build AI. You'll gain hands-on experience with cutting-edge technologies and methodologies.
Flexible Commitment
We understand you're a student first. We offer flexible hours and remote work options that fit around your academic schedule.
Community of Peers
Join a community of like-minded students passionate about AI. Collaborate, learn, and grow together with talented peers.
Open Positions
AI Product & Full-Stack Engineer
Own end-to-end development of TAI Teens across mobile and web: connect the model inference API, the billing system (payments / redeem codes) and the interface into one working loop.
What You'll Do:
- Ship the TAI Teens web and mobile surface end to end
- Wire inference, billing and UI into one loop
- Take ownership of uptime and payment correctness
We're Looking For:
- Fluent in at least one backend language (Python / Node.js) and a frontend framework (React / Vue)
- Have shipped a complete consumer product on your own
- Stay calm and fix things fast when a server falls over or a payment callback fails
Alignment & Data Strategy Researcher
Design the dialogue data pipeline and lead SFT and LoRA finetuning, with safety guardrails, persona alignment, and the long-context problems of logical idling and forgetting.
What You'll Do:
- Design and run the SFT and LoRA pipeline
- Align safety guardrails and persona
- Fix logical idling and forgetting in long context
We're Looking For:
- Know the post-training stack (SFT / DPO / RLHF) and tools like HuggingFace / PEFT
- Extremely sensitive to data: can spot template collapse from a few hundred conversations
- Can script automated cleaning and synthetic data generation
AI Systems & Inference Engineer
Build and maintain the inference infrastructure: quantise 4B-class models, speed up inference with vLLM / llama.cpp / MLX, and make them run well on Mac and on-device.
What You'll Do:
- Own the inference stack and deployment
- Quantise and accelerate 4B-class models
- Find the cheapest path to smooth concurrency
We're Looking For:
- Understand computer architecture and GPU / Apple Silicon characteristics
- CUDA or Metal experience is a strong plus
- Have worked with elastic compute scheduling on a cloud platform
Multimodal Video / Audio Researcher
Explore joint video and audio generation, and solve copy degeneration and closed-loop state explosion in long-horizon video.
What You'll Do:
- Research next-generation joint video/audio architectures
- Attack degeneration in long-horizon generation
- Validate ideas cheaply with synthetic data
We're Looking For:
- Deep understanding of diffusion models, autoregressive video generation and world models
- Comfortable with optical flow, event alignment and continuous state estimation
- Can validate architectural ideas with small synthetic datasets under tight compute
Post-Training Engineer
Squeeze maximum capability out of open bases like Qwen and Llama through data refinement, and find the path from 1B-4B models toward large-model quality.
What You'll Do:
- Refine data and post-train open bases
- Push 1B-4B models toward larger-model quality
- Keep experiment cost as low as possible
We're Looking For:
- Deep knowledge of LoRA / QLoRA and hands-on fixes for catastrophic forgetting and overfitting
- Can tailor training to a target task (code, reasoning)
- Obsessive about compute cost: best convergence for the smallest experiment
Frontier Researcher
Look past current scaling laws for the next paradigm: continual learning, self-improving models, recurrent architectures, neuromorphic computing.
What You'll Do:
- Explore paradigms beyond current scaling laws
- Design and run long-horizon experiments
- Question the assumptions everyone else accepts
We're Looking For:
- Strong critical thinking and the nerve to question mainstream deep learning
- Solid grounding in mathematics, physics or cognitive science
- Willing to take big research risk on hard problems for a long time
- A self-built, ambitious, still-imperfect experimental project counts for a lot
Don't see a perfect match but still interested in joining us?
Get In TouchApplication Process
Apply
Submit your application with your resume and any relevant projects
Initial Chat
We'll have an informal conversation to learn about your interests and experience
Technical Discussion
We'll discuss technical concepts relevant to the role
Offer
If it's a good fit, we'll extend an offer to join our team
