Fellowship · Remote (India)
AI Reasoning & Evaluation Intern
Cehpoint is hiring for this role: fellowship, Remote (India), ₹15,000 – ₹24,000/month + Full-Time PPO Option. Build benchmark datasets, evaluate chain-of-thought accuracy, and verify mathematical logic deduction traces for frontier AI models.
About the role
Evaluation is the engine of AI progress. As a Reasoning & Evaluation Intern, you will verify whether agents genuinely deduce correct solutions or merely hallucinate plausible-looking steps, ensuring our reasoning engines meet enterprise accuracy bars.
Assist in building the testing pipelines that evaluate our reasoning agents. You will curate reasoning problems, score automated deduction traces, and benchmark model accuracy across complex logic and coding challenges.
What you will do
- Assemble and annotate high-difficulty benchmark problems spanning formal logic, math, and software architecture.
- Execute automated test runs across frontier models and analyze reasoning trace failure points.
- Implement scoring heuristics that verify intermediate reasoning lemmas.
- Visualize accuracy trends and token efficiency metrics for engineering leadership.
- Support senior researchers in prompt optimization and error categorization.
What you should know
- Strong foundation in discrete mathematics, logic, or computer science theory.
- Familiarity with Python for data manipulation and test automation.
- Understanding of reasoning techniques: Chain-of-Thought, Self-Consistency, and verification.
What you should be able to do
- Proficiency with Python, Pandas, and JSON data structures.
- Sharp analytical eye capable of spotting subtle logic flaws in model outputs.
- Disciplined data management and documentation skills.
How you work
- A love for logic puzzles, competitive programming, or mathematical problem solving.
- Reluctance to accept unsubstantiated assertions without empirical verification.
- High ambition to develop deep expertise in machine learning evaluation.
Skills we check
What you get
- Paid stipend (₹15,000–₹24,000/month) with performance bonuses.
- Direct PPO track into Full-Time Synthetic Reasoning Architect.
- Deep practical immersion in frontier AI reasoning architectures.
- Flexible remote working conditions.
How we select for this role
We read every application ourselves. Here is what happens after you apply.
Apply with a short form
Paste a link to your CV and answer five questions in your own words. No sign-in and no file upload. About fifteen minutes.
Integrity check
When you press Check my answers, an automated check confirms the answers were typed by you. You see the result straight away, before you submit.
Choose your project date
If no cheating is detected you are shortlisted, and we email you a link to choose the day your project starts.
5-day practical project
On that day our team contacts you with the brief: real work, 5 working days, reviewed by an engineer. For this role: Construct an automated evaluation suite that tests a model on 30 multi-step logic problems, scoring intermediate steps and computing verification precision (strictly 5 working days).
Offer
If your work solves the problem well, you receive your offer letter by email and in the portal.
Other open roles in this team
- AI Agent Cognition ResearcherAI engineering & research · Full-time · Remote (India)₹16,00,000 – ₹28,00,000 / year + Research Grant
- AI Fast-Decision Systems ResearcherAI engineering & research · Full-time · Remote (India)₹12,00,000 – ₹22,00,000 / year + Research Grant
- AI Memory & Knowledge Graph EngineerAI engineering & research · Full-time · Remote (India)₹11,00,000 – ₹19,00,000 / year
- AI Reasoning Systems ArchitectAI engineering & research · Full-time · Remote (India)₹15,00,000 – ₹25,00,000 / year + Performance Bonus
- AI Supply Chain & Data Ops EngineerAI engineering & research · Full-time · Remote (India)₹12,00,000 – ₹20,00,000 / year
- AI Workflow & Autonomous Agent ArchitectAI engineering & research · Full-time · Remote (India)₹10,00,000 – ₹18,00,000 / year + Equity Option