Research Scientist, Applied AI

Build the core agent behaviour behind a revenue-focused AI agent, from the agent harness and evals to fine-tuning and reinforcement learning, and take hard research problems from hypothesis to production. 

• Key Responsibilities 
• Own and advance the agent harness: the core loop, tools, environment, guardrails, and model behaviour that power the agent. 
• Improve how the agent selects tools, builds context, coordinates across sub-agents, and handles long, multi-step tasks, through systematic experimentation. 
• Drive model selection across tasks, balancing answer quality, latency, and cost. 
• Design and build evaluations for complex revenue tasks such as deal health, risk, and pipeline questions, including benchmarks, regression detection, and automated eval pipelines. 
• Build eval datasets and grading methods for tasks that have no clean ground truth. 
• Build scoring and prediction models, such as deal health scores from call transcripts and emails, cross-deal pattern mining, and explainable risk signals. 
• Lead fine-tuning work as the platform moves from harness-heavy work into deeper model training.
• Research and design reinforcement learning (RL) approaches that improve agent behaviour. 
• Improve context engineering and data quality through entity resolution, information extraction, and knowledge and context graphs. 
• Own ambiguous research problems end-to-end: form hypotheses, design experiments, build the training, eval, and grading infrastructure, and ship results to production. 
• Work closely with the US-based Head of AI and the applied AI team in India. 

• Qualifications
• 5+ years of experience in applied AI research or applied science, at a senior level. 
• MS or PhD in Computer Science, Mathematics, Statistics, or another quantitative discipline. 
• Hands-on experience building and improving agentic systems and agent harnesses, including tool use, multi-step reasoning, and context management. 
• Hands-on experience with LLMs or large-scale deep learning systems, including training, fine-tuning, and RLHF. 
• A clear understanding of why models behave the way they do, not only how to use them. 
• Experience designing eval datasets or frameworks, especially for tasks without clean ground truth.
• Proven ability to scope and drive ambiguous research with minimal guidance. 
• Publications at top-tier peer-reviewed conferences or journals are preferred. 
• Comfortable working with a US-based leadership team across time zones.
• Benefits & Perks 
• Competitive salary and equity at Series A, ahead of the next funding round. 
• A remote role, with the team meeting in person in Bangalore. 
• Meaningful scope from day one: you will own core agent capabilities, not just support them. 
• Direct work with the Head of AI and a tight feedback loop with founders and product. 
• A sharp, low-ego team that moves fast and communicates directly. 
• A chance to shape an AI-first product at a pivotal stage, as it moves from harness work into fine-tuning and RL. 
   

Must-have skills

Agentic Systems, Agent Harness Design, model finetuning

Good-to-have skills

LLM, Applied AI, AI research