Job Description

Job Summary

AI Engineer for Bias Assessments

The AI Engineer for Bias Assessments serves as HCLTech’s central technical specialist for conducting bias and fairness assessments, investigating potential root causes, and recommending and validating mitigation measures. The role combines hands-on statistical analysis and AI engineering with practical knowledge of responsible AI, algorithmic auditing, explainability, and data governance.

The engineer assesses HCLTech-developed and third-party AI systems used in HCLTech’s internal operations, including models, AI agents, decision-support tools, and AI-enabled workflows. The role identifies potential unfair impacts and disparate performance across relevant population groups and produces documented, reproducible evidence to support governance decisions.

The engineer works alongside LoB testers, engineering teams, and data scientists, providing specialist methods, tools, training, and technical review. LoB teams retain responsibility for their assigned testing and implementation activities.

Key Responsibilities

Bias and Fairness Assessments

• Plan and conduct assessments across the AI lifecycle, including before deployment, after material changes, periodically during operation, and following relevant incidents.
• Define assessment scope, intended use, affected populations, potential harms, data requirements, metrics, and acceptance criteria with system owners and relevant stakeholders.
• Assess classification, regression, ranking, recommendation, and decision-support systems; adapt evaluation methods for generative AI and agentic workflows.
• Select, calculate, validate, and interpret appropriate measures, including selection and scoring rates, impact ratios, demographic parity, equal opportunity, equalized odds, error-rate disparities, calibration, and intersectional performance.
• Evaluate statistical uncertainty, confidence intervals, effect sizes, sample sufficiency, and practical significance. Identify when available evidence is insufficient to support a conclusion.
• Develop reusable assessment scripts, notebooks, test suites, and reporting templates using Python and appropriate fairness and statistical tools.

Data Evaluation and Root-Cause Investigation

• Determine whether training, validation, test, benchmark, production, and reference datasets are suitable for each assessment.
• Evaluate data provenance, lineage, representativeness, missingness, imbalance, labeling quality, measurement error, data leakage, and potential proxy variables.
• Identify relevant protected characteristics and intersectional groups, working with Legal and Privacy to confirm appropriate collection, access, use, and handling of sensitive data.
• Investigate potential sources of disparities across data collection, labeling, feature transformations, model objectives, thresholds, prompts, retrieval sources, agent interactions, and human decision processes.
• Use subgroup analysis, error analysis, controlled comparisons, sensitivity analysis, and explainability techniques such as SHAP or LIME to investigate potential causes. Recognize that feature attribution alone does not establish causation or demonstrate fairness.

Generate or evaluate synthetic data where appropriate for scenario testing or underrepresented cases, documenting its assumptions, limitations, and suitability.

Mitigation and Validation

• Recommend practical mitigation options, including data remediation, feature or model changes, fairness constraints, threshold adjustments, prompt or retrieval improvements, human review, workflow controls, monitoring, and use restrictions.
• Explain trade-offs involving fairness, accuracy, privacy, reliability, and operational performance.
• Work with accountable technical and business teams to establish remediation plans with owners, actions, milestones, target dates, acceptance criteria, and validation methods.
• Reassess systems after remediation to determine whether disparities have improved and identify residual risks or unintended effects. Recommend ongoing monitoring for changes in data, system behavior, users, and affected populations.
• Escalate material, unexplained, or recurring disparities and overdue corrective actions through HCLTech’s AI governance process.

LoB Enablement and Audit Support

• Serve as the central technical expert supporting LoB testers through assessment planning, troubleshooting, peer review, and interpretation of findings.
• Develop practical guidance and train internal teams to conduct consistent, reproducible assessments.
• Prepare assessment reports, stakeholder summaries, and audit evidence covering scope, datasets, methodology, results, limitations, root-cause findings, remediation, and residual risks.
• Explain findings clearly to Engineering, Data Science, business owners, governance, Legal, Privacy, Risk & Compliance, and leadership.

Skill Requirements

• Demonstrated hands-on experience independently conducting bias and fairness assessments, with the ability to explain dataset selection, metric rationale, statistical limitations, root-cause investigation, mitigation options, and validation results.
• Strong Python and SQL proficiency, including practical experience with pandas, NumPy, scikit-learn, statistical libraries, and reproducible analytical workflows.
• Hands-on experience with AI Fairness 360, Fairlearn, or equivalent tools, with the ability to explain and validate the underlying calculations.
• Strong understanding of fairness metric families, their assumptions and limitations, and trade-offs between different fairness objectives.
• Strong statistical skills, including sampling, confidence intervals, hypothesis testing, effect sizes, small-sample limitations, and intersectional analysis.
• Experience selecting appropriate datasets, identifying data-quality problems and proxy variables, and investigating potential sources of bias.
• Working knowledge of explainability techniques such as SHAP and LIME and their limitations.
• Ability to translate assessment findings into implementable technical and operational improvements.
• Experience producing evidence suitable for governance reviews or external audit scrutiny and explaining methodology to reviewers.
• Ability to advise and train internal testers and coordinate work across technical and governance teams.
• Working knowledge of protected characteristics, discrimination risks, responsible AI, data protection, and AI governance, with the judgment to seek jurisdiction-specific advice.
• Clear written and verbal communication, sound technical judgment, attention to detail, and willingness to escalate difficult findings.
• Degree in computer science, data science, statistics, mathematics, engineering, or a related discipline, or equivalent practical experience.

Other Requirements