About Deccan AI

Deccan AI is a high-growth, venture-backed AI model training and evaluation company headquartered in the Bay Area. Founded by alumni of IIT Bombay, IIM Ahmedabad, and ex-Google , we partner with the world’s top AI frontier labs including Google DeepMind, Snowflake , and several cutting-edge research groups. We are backed by Prosus Ventures , and our India office is based in Hyderabad.

We’re not just participating in the AI race we’re building the infrastructure that powers it.

With 1M+ global experts, advanced automation, and vertically integrated platforms, we deliver the gold-standard data that world-class AI models rely on. The AI data annotation market is exploding set to quadruple by 2032. The opportunity? Massive, and you can help define the future

**.

Backgrou**

ndThe ML Researcher plays a critical role in keeping the organization at the forefront of AI innovation through deep, end-to-end research. This role is focused on identifying emerging research directions, designing novel AI benchmarks, and creating high-quality evaluation datasets aligned with both cutting-edge research and organizational prioritie

s.The ML Researcher bridges the gap between theoretical research and practical application by translating insights from the latest AI papers into actionable benchmarks and evaluation frameworks. The role emphasizes deep research, synthesis of academic and industry findings, and the design of solutions that meaningfully advance AI systems—particularly in agent frameworks, machine learning models, and emerging AI paradigm

**s.

Purpose of the R**

oleThe primary objective of the ML Researcher is to conduct comprehensive AI research and desi gn novel, scalable, and real-world–relevant benchmarks and evaluation datas ets. The role aims to push beyond existing evaluation practices by defining new criteria that better measure the capabilities and limitations of modern AI syste

ms.

Role Over

viewThis role focuses on evaluat ing code generation and agent-driven coding workf lows, including planning, debugging, and multi-step development ta

sks.Key Responsibili

• tiesResearch code generation models and agentic coding syst
• ems.Design benchmarks for algorithmic coding, software tasks, and workfl
• ows.Propose evaluation datasets covering real-world coding scenar
• ios.Define metrics for correctness, robustness, maintainability, and workflow succ
• ess.Produce benchmark documentation for pipeline te

ams.Delivera

• blesCoding and agentic coding benchmark propo
• salsCode evaluation data
• setsMetrics for code quality and agent perform

anceRequired Sk

• illsStrong programming and software engineering backgr
• oundResearch experience in code models or program synth
• esisAbility to design realistic coding evaluat

ions