Python Developer

Python Developer – Hallucination Detection Platform

Moleculyst Ventures Private Limited is seeking a Python Developer to contribute to our cutting-edge AI infrastructure for detecting and analyzing hallucinations in large language models (LLMs). This role focuses on developing software systems to identify unreliable outputs from LLMs by analyzing inference-time data, model computations, and internal signals.

The ideal candidate will design, build, and maintain modular Python systems for running open-weight language models, processing inference data, and creating backend services. You will collaborate closely with AI/ML researchers to translate experimental methods into scalable, reliable software solutions.

Key Responsibilities:

• Develop and maintain clean, modular, and testable Python systems for LLM infrastructure.
• Build pipelines for capturing, processing, and storing inference-time data, including model activations, tensors, metadata, prompts, and outputs.
• Design APIs and backend services for model evaluation and hallucination detection.
• Work with large-scale numerical datasets, handling tensors, prompts, and generated outputs efficiently.
• Integrate databases (SQL/NoSQL) and storage systems for optimized retrieval and analysis.
• Write comprehensive unit and integration tests to ensure correctness and reproducibility.
• Profile and optimize performance-critical Python code for efficiency.
• Document system architecture, APIs, data formats, and engineering decisions.
• Collaborate with researchers to convert experimental code into robust, production-ready software.

Qualifications:

• Strong Python programming skills with expertise in object-oriented programming, data structures, and software design.
• Experience with NumPy and PyTorch, preferably with internals of transformer-based models.
• Familiarity with backend development, APIs, and database systems (SQL/NoSQL).
• Experience with Git, testing, debugging, and standard software development practices.
• Comfort with Linux, command-line environments, and large/complex datasets.
• Ability to independently solve technical problems and optimize performance.
• Strong communication skills for remote collaboration.
• Bonus: Experience with Hugging Face Transformers, Docker, cloud infrastructure, and GPU computing (CUDA).

What You’ll Work On:

A typical workflow includes: Prompt → LLM Inference → Data Capture → Signal Processing → Hallucination Detection → Storage/API. This role bridges Python engineering, ML infrastructure, and LLM interpretability, requiring ownership, problem-solving, and close collaboration with the research team.