Synthetic Reasoning & Evaluation Intern

Cehpoint; India; Remote; INR 167,000 - 268,000; Full time

Benefits offered by this job

Paid stipend

Flexible remote work

PPO track into full-time role

Job summary

Reasoning & Evaluation Internship — we are seeking a keen intern to help verify whether agents deduce correct solutions or hallucinate steps, ensuring enterprise-grade reasoning accuracy.

You will work on Python-based benchmarks, data analysis with Pandas, and JSON tooling, contributing to automated evaluation suites and progress visualization for leadership.

Qualifications

• 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.
• Proficiency with Python, Pandas, and JSON data structures.
• Disciplined data management and documentation skills.

Responsibilities

• 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.

Skills

Discrete mathematics

Logic

Computer science theory

Python programming

Data manipulation

Pandas

JSON data structures

Data analysis

Testing automation

Analytical thinking

Tools

GitHub

Job description

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.

Python Mathematical Logic Reasoning Benchmarks Prompting Techniques Data Analysis Pandas GitHub

Section 02

• 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.

Section 03 Mandatory Foundational Knowledge

• 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.

Section 04 Mandatory Practical Skills & Architecture

• 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.

Section 05 Problem Solving, Execution Rigor & Curiosity

• 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.

Section 06 Practical evaluation benchmark

5-day live technical evaluation milestone

5-Day Practical Milestone: 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).

Institutional hiring protocol: candidates who clear resume screening take a live practical milestone of strictly 5 working days. Verifiable completion and code audit by your assigned engineering mentor is the sole prerequisite for the official offer letter.

Section 07 Compensation, Total Rewards & Advancement

• 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.

Dedicated inbox for this role

Questions, private repository links or portfolio references for this opening route straight to the engineering leads reviewing it.