TELUS Digital is seeking research-level mathematicians to contribute to an advanced mathematical reasoning benchmark for a leading AI research organization.
As a Complex Math Data Specialist, you will design a single original, research-level mathematical problem intended to probe the limits of current AI reasoning systems. The problem should sit beyond the capabilities demonstrated by current frontier models under the programme's evaluation protocol and must have a precise, machine-checkable answer.
This is problem authoring, not proof formalization. You will not be asked to produce a machine-checked proof. Instead, you will create one rigorous mathematical challenge, provide its exact answer and solution, and supply the computational materials required to independently verify the result.
The project is particularly suited to researchers with deep expertise in areas such as number theory, combinatorics, algebraic geometry, analysis, topology, differential geometry, PDEs, harmonic analysis, category theory, or mathematical physics.
What You'll Do
• Design one original research-level problem in your area of mathematical expertise. The problem must be new and unpublished. You may build on existing mathematical ideas, but the resulting problem must be sufficiently transformed or combined that its source is not readily identifiable.
• Make the problem completely self-contained. Include all definitions, conventions, non-standard notation and hypotheses required to solve it. The evaluator will not have access to supplementary papers or references at evaluation time.
• Define one exact answer and answer type. Accepted formats may include a Python integer, sympy.Expr, sympy.Matrix, tuple of integers, or another explicitly supported machine-checkable format. Approximate answers do not receive credit.
• Make the problem resistant to guessing. The probability of obtaining the correct answer without performing the substantive mathematical work should be below the programme's threshold. Small integers, familiar constants, simple counts and yes/no answers are generally unsuitable.
• Provide a complete solution write-up, presenting the key mathematical idea first, followed by the reasoning and any required computation.
• Provide reproducible code that computes or verifies the answer and runs in under approximately 60 seconds on ordinary hardware.
• Provide the required reference-answer function in the programme's submission format.
• Evaluate your own problem using the programme's testing harness, including repeated independent model attempts, a statement-only guessing attack and contamination/source-identification checks.
• Submit evaluation logs and the required difficulty metadata, including subject area, techniques used, estimated expert hours to discover the key idea, estimated execution time once the idea is known, and whether programming is required.
Mandatory Qualifications
• Active mathematical research at doctoral level or beyond, such as a senior PhD student with supervisor endorsement, postdoctoral researcher, early-career faculty member or professor.
• Demonstrated depth in a mathematical research area, supported by published or ongoing research. Relevant areas include:
• Number theory
• Combinatorics
• Algebraic geometry
• Analysis
• Topology
• Differential geometry
• Partial differential equations
• Harmonic analysis
• Category theory
• Mathematical physics
• Programming capability sufficient to compute and verify your own result, preferably using Python. Available tools include NumPy, SymPy, SciPy, NetworkX, gmpy2, mpmath and galois.
• Strong mathematical rigour, particularly the ability to state all assumptions and definitions explicitly and identify potential gaps under adversarial review.
• Publication-standard mathematical English.
• Ability to revise the submission following independent mathematical review.
• Willingness to work under strict confidentiality, including compliance with programme NDA requirements and restrictions on external testing or disclosure of the submitted problem.
Qualification path
Nice to Have
• Previous experience writing mathematical benchmarks, olympiad or competition problems
• Experience with AI reasoning evaluation
• Computational algebra or number theory experience using Sage, Magma or PARI/GP
• Formalization experience with Lean or Coq
• Research in comparatively underrepresented areas such as differential geometry, PDEs, harmonic analysis, category theory or mathematical physics
Job details
Location
Chennai, India
Job type
Part-time
Work arrangement
Remote
Experience
Entry level
Category
Data Entry
About the company
Telus Digital
Employer profile
TELUS Digital is at the forefront of creating and enhancing the world's data to enable better AI via human intelligence. By providing enriched data through a skilled team of AI specialists and a managed AI Community of over one million crowd contributors, we empower generative AI (GenAI), computer vision, and natural language processing models. Utilizing proprietary AI-assisted, quality-optimized technology, we ensure efficient data collection, accurate labeling, and precise fine-tuning. Our solutions improve AI systems in multiple applications, such as advanced smart products, improved search results, expanded speech recognition, more human-like bot interactions, and more.
