Senior AI Researcher, Distributed Large Language Model Training

About Templar

Templar is a pioneering community of researchers and engineers pushing the boundaries of permissionless pretraining. We are advancing the frontier of distributed AI development by enabling collaborative training across diverse computational resources without traditional centralized constraints. Our work is detailed in our recent research: Incentivizing Permissionless Distributed Learning of LLMs.

Overview

We're seeking an exceptional AI researcher to advance the frontier of distributed training for Large Language Models. You'll lead cutting-edge research efforts exploring novel approaches to distributed LLM pretraining, with a focus on communication optimization, incentivization mechanisms, and trust frameworks in permissionless environments. Your work will directly contribute to scaling our proven approach and developing the next generation of collaboratively trained language models.

Responsibilities

• Distributed LLM Training Research
• Drive foundational research in decentralized optimization algorithms specifically for large language model pretraining
• Develop innovative gradient aggregation approaches that maintain training stability across heterogeneous nodes during LLM training
• Research efficient distributed learning mechanisms that minimize communication overhead for transformer architectures
• Design novel trust and quality assessment frameworks for distributed LLM training contributions

Investigate scaling laws and convergence properties specific to distributed language model training

System Architecture & Optimization for LLMs

• Conceptualize scalable architectures for decentralized LLM training across the full pretraining lifecycle
• Research compression and sparsification techniques optimized for transformer gradient patterns and network bandwidth requirements
• Develop synchronization mechanisms resilient to node failures during long-running LLM training processes
• Create adaptive optimization techniques suitable for varied computational environments in language model training
• Design efficient data loading and preprocessing pipelines for distributed LLM training

Research Leadership

• Lead a multidisciplinary team investigating distributed LLM training systems
• Establish research roadmaps aligning theoretical advances in distributed optimization with practical LLM training implementations
• Collaborate with academic and industry partners to advance distributed language model training
• Publish findings in top-tier conferences and journals focused on machine learning and distributed systems
• Mentor junior researchers and contribute to the broader research community

Required Qualifications

• PhD/MSc in Computer Science, Machine Learning, or related field
• Strong research background in distributed systems, machine learning optimization, or large-scale model training
• Deep understanding of transformer architectures and large language model training dynamics
• Expertise in modern deep learning frameworks (PyTorch, JAX) and distributed computing systems
• Demonstrated ability to translate theoretical research into practical implementations for large-scale systems
• Excellent written and verbal communication skills with experience presenting research findings

Preferred Experience

• Research publications in distributed learning, federated optimization, or large language model training
• Hands-on experience with LLM pretraining, including data preparation, training optimization, and evaluation
• Background in information theory, compression, or signal processing applied to neural networks
• Contributions to open-source machine learning projects, particularly those related to distributed training or LLMs
• Experience with permissionless or incentivized distributed computing systems

Key Research Areas

• Asynchronous and semi-synchronous distributed optimization for transformer training
• Gradient compression and efficient communication protocols tailored to LLM parameter patterns
• Trust mechanisms and contribution quality assessment in permissionless LLM training
• Specialized techniques for distributed domain adaptation of language models
• Scalable architectures for heterogeneous computing environments in LLM training
• Incentivization mechanisms for sustainable distributed LLM pretraining

Potential Short-Term Objectives (0-6 Months)

Conduct comprehensive analysis of current distributed LLM training approaches, particularly in permissionless settings

Benchmark and evaluate existing distributed training methods on language model tasks

Identify key research directions for improving communication efficiency, robustness, and verification in distributed LLM training

Develop experimental frameworks for rigorously comparing distributed optimization strategies for transformer models

Design and prototype trust-based contribution evaluation mechanisms for LLM training participants

Collaborate with engineering teams to address practical challenges in scaling our training approach

Potential Long-Term Objectives (6+ Months)

• Pioneer novel gradient compression and aggregation techniques optimized for transformer architectures
• Develop scalable architectures supporting LLM training across heterogeneous node capabilities and trust levels
• Create frameworks for domain-specific LLM adaptation in decentralized environments
• Establish theoretical foundations for trust-weighted learning with unverified participants in LLM training
• Lead research initiatives resulting in breakthrough publications and open-source contributions to the distributed LLM training community
• Drive the development of Templar's next-generation distributed language models

Impact

Your research will directly advance Templar's mission to democratize AI development through permissionless, distributed LLM training. You'll help create more efficient, accessible, and scalable distributed training systems that enable collaborative development of state-of-the-art language models. Your work will contribute to a future where cutting-edge AI capabilities can be developed through community collaboration rather than concentrated computational resources, fundamentally changing how large language models are created and improved.

How to Apply

Send your application materials to: [email protected]

As a Bittensor Trading Engineer, you will be a critical member of the Yuma investment team and at the forefront of trading and portfolio execution within the Bittensor ecosystem. You will actively manage and optimize on-chain trading strategies across tokens. This role uniquely combines quantitative investment expertise with hands-on coding skills, allowing you to build, deploy, and oversee algorithmic trading models that maximize returns within the Dynamic TAO (DTAO) framework.

You will work closely with the Yuma Investment Lead to translate investment strategy into code, ensuring effective trade execution, liquidity optimization, and risk management. You will leverage smart contract interactions, trading bots, and automated portfolio rebalancing techniques to maintain a high-performing, data-driven investment approach.

KEY RESPONSIBILITIES :

Algorithmic Trading & Execution

Write, manage, and oversee the execution of dynamic TAO trading strategies across a portfolio of tokens. Develop, deploy, and optimize algorithmic trading scripts to automate trading decisions and portfolio rebalancing. Implement and maintain smart contract integrations for automated trade execution, staking, and liquidity provisioning. Monitor on-chain order flow, token velocity, and market signals to adjust trading strategies in real time. Ensure efficient execution of staking, validator participation, and yield-maximizing strategies. Portfolio Construction & Investment Strategy

Work closely with the Yuma Investment Lead to execute on-chain portfolio construction decisions. Analyze liquidity, slippage, and order book depth to optimize trade execution across multiple subnet tokens. Utilize quantitative and statistical models to assess risk-adjusted returns and improve trade execution strategies. Implement risk management frameworks, including dynamic hedging strategies and portfolio optimization techniques. Develop custom dashboards and trading analytics tools to monitor portfolio performance, P&L, and market impact. Research & Market Intelligence

Conduct deep technical and economic research on Bittensor, TAO, and subnet token markets. Assess tokenomics models, staking incentives, and liquidity incentives for investment opportunities. Track on-chain metrics and market data using Dune Analytics, Nansen, or custom-built data pipelines. Identify emerging trends in decentralized AI, validator economies, and subnet activity that influence investment decisions. Stay ahead of regulatory developments, smart contract risks, and security vulnerabilities in automated trading systems. Technical & Risk Management

Ensure trade execution reliability and efficiency, maintaining uptime and accuracy in automated systems.

• Implement automated risk controls, including circuit breakers, position limits, and execution safeguards.
• Work with cross-functional teams to optimize system architecture, API integrations, and decentralized execution pipelines.
• Stay ahead of Bittensor subnet developments, ensuring trading strategies remain adaptive and competitive.

QUALIFICATIONS:

Proficiency in coding languages: Python, Rust, Solidity, SQL, or other relevant programming languages for on-chain trading execution. Experience in developing trading algorithms, automated execution strategies, or DeFi smart contract interactions.

• Strong understanding of blockchain protocols, validator dynamics, and decentralized exchanges (DEXs).
• Hands-on experience with smart contract interactions, oracles, and automated trading bots. Technical fluency in smart contract interactions, blockchain data, and automated trading mechanisms (Python, SQL, Solidity preferred).
• Deep understanding of tokenomics, staking incentives, validator economics, and liquidity provisioning.
• Experience in on-chain financial modeling, risk assessment, and portfolio optimization.
• Strong ability to translate investment theses into executable, algorithmic trading strategies Preferred Qualifications

3+ years of experience in digital asset trading, quantitative finance, hedge funds, or crypto investment strategies. Familiarity with staking-as-a-service, perpetual futures, structured DeFi products, and liquidity mining strategies. Experience with Bittensor, TAO, and subnet token staking mechanisms. Understanding of DeFi, perpetual futures, options trading, and structured crypto products. Prior experience developing algorithmic trading models for digital asset hedge funds or quant desks. WHAT WE OFFER:

An opportunity to thrive in a dynamic, cutting-edge environment at a rapidly scaling company led by experienced industry leaders An innovative learning environment where you can immerse yourself in the latest technologies, contribute to building a transformative new industry, and make a meaningful impact.Competitive base salary, bonus and incentive compensation

• Unlimited PTO / Flexible time off - work with your leader to take time off when you need it
• Professional development budget with flexibility for personal and professional growth
• Outstanding health insurance for employee, partner and dependents
• Life insurance, short-term & long-term disability coverage
• 401K plan with company contribution
• Flexible spending programs for medical and dependent care
• Paid parental leave