ARMMAN, Technical AI Analyst
About the portfolio organization
Founded in 2008, ARMMAN’s mission is to enable healthy pregnancy, safe delivery and safe childhood for women and children in India. ARMMAN leverages mHealth to create cost-effective and scalable systemic solutions to improve access of pregnant women and mothers to preventive information and services and train health workers to reduce maternal and child mortality and morbidity. Our programs have reached over 60M women and over 400k health workers across 27 states to date.
About the Fellowship role
The AI & Data Integrity Fellowship is a high-impact, technical role designed for someone who sits at the intersection of Generative AI and Data Engineering. While many focus solely on "chatting" with AI, this role is about industrializing those interactions. The Fellow will be responsible for ensuring that our LLM outputs are safe, accurate, and optimized for cost/latency, while simultaneously building the "trust layer" in our MySQL databases to ensure the AI isn't learning from—or reporting—garbage data.
Employment: Full-time, one-year Fellowship
Starting Date: July 2026
Key responsibilities
LLM Optimization & Safety
- Prompt Engineering & Testing: Design, version control, and A/B test system prompts to improve response accuracy and tone.
- Hallucination Mitigation: Implement "Chain of Thought" or "Self-Consistency" prompting techniques to reduce factual errors.
- Guardrail Development: Stress-test the model against prompt injections, PII leaks, and toxic outputs.
- Benchmarking: Create "Golden Datasets" to quantitatively score LLM performance (e.g., scoring responses on a scale of 1-5 for accuracy).
Data Quality & MySQL Engineering
- Validation Layer: Build SQL scripts or Python triggers to ensure data entering the system meets specific formats and business logic.
- Quality Check Layer: Develop automated checks for null values, duplicates, and referential integrity within MySQL.
- RAG Readiness: Ensure that the data being retrieved for the LLM is "clean" and indexed properly for high-speed retrieval.
Requirements
Experience and education:
Technical Skills
- SQL Proficiency: Deep understanding of MySQL (Joins, Indexing, Constraints, and stored procedures).
- Python for AI: Experience with libraries like LangChain, LlamaIndex, or OpenAI SDK.
Testing Mindset: Familiarity with evaluation frameworks (e.g., RAGAS, DeepEval) or a strong background in manual QA. - Data Modeling: Basic understanding of how schema design affects data quality.
Soft Skills
- Analytical Rigor: The ability to spot a "hallucination" that looks like a fact.
- Academic Background: Currently pursuing or recently graduated with a degree in Computer Science, Data Science, or a related field.
- Ethical Compass: A strong understanding of AI safety and the importance of data privacy.
Must Haves:
- Portfolio or GitHub showing at least one RAG (Retrieval-Augmented Generation) project.
- Experience with MySQL.
- Proven ability to write clean, modular Python code.
- 1-2 years of relevant experience
- Authorization to work in India
