> Markdown version of [/jobs/ext/4749902-principal-software-engineer-it](https://www.wearedevelopers.com/jobs/ext/4749902-principal-software-engineer-it). Every page supports `.md` or `Accept: text/markdown`. Links point to the HTML versions so they work for humans too. Agent guide: [/agents.md](https://www.wearedevelopers.com/agents.md). --- # Principal Software Engineer - IT - **Company:** Dell Technologies Inc. - **Location:** United States - **Experience:** Expert - **Contract:** Internship / Graduate position - **Skills:** LangGraph Framework, Artificial Intelligence, Continuous Integration, Python (Programming Language), Machine Learning, Open Web Application Security, Software Engineering, Large Language Models, Multi-Agent Systems, Langfuse — LLM Observability and Analytics Platform - **Published:** November 16, 2026 - **Apply:** https://www.jobmonkeyjobs.com/career/28187641/Principal-Software-Engineer-It-Any-Bangalore-1312 ## About the Role * 8+ years in software engineering, quality engineering or machine learning evaluation, with systems shipped to production. * Hands-on construction of evaluation harnesses in Python: dataset design, test orchestration, scoring implementation and result reporting for non-deterministic systems. * Ability to build measurable analytics of models and agentic tools, using LangFuse (or equivalent LLM observability tooling) and embedding-based comparison techniques. * Statistical literacy: sample size, variance across runs, and the difference between a real change in behaviour and noise, applied with Python data tooling at volume. * Experience owning a quality gate or release decision, including declining a release and holding that position under schedule pressure. Desirable Requirements * Experience evaluating AI systems within a security domain, and familiarity with agent orchestration frameworks such as LangGraph to understand what is being evaluated. * Familiarity with the OWASP Top 10 for LLM Applications and the NIST AI Risk Management Framework, ideallyin an audited or regulated environment. ## Description * You will own the release gate for AI-native security capabilities, deciding what evidence is required before a workflow can go live. * You will build evaluation harnesses that measure model and agent behaviour objectively, not by sampling outputs and forming an impression. * You will turn evaluation into a repeatable, measurable analytics practice so that every workflow after the first costs less to evaluate than the one before it. You will: * Define and record acceptance criteria for each capability with the functional owner before evaluation begins, and specify the evidence, ground truth and coverage required to support the claim. * Build and maintain evaluation harnesses in Python covering dataset construction, test orchestration, scoring and reporting, and integrate them as automated quality gates in CI/CD. * Design measurable analytics for models and agentic tools - quantifiable metrics for accuracy, consistency across repeated runs, confidence-versus-accuracy calibration, latency and cost - using tooling such as LangFuse for trace-level evaluation and embedding-based similarity scoring for output comparison. * Own the release gate: apply the criteria to the evidence, record the result, and block releases that do not clear the threshold, with thresholds set proportionate to the consequence of an incorrect agent action. * Run ongoing regression evaluation after release to detect behavioural drift from model, prompt, data or upstream changes, and report results to engineering and functional stakeholders in decision-ready terms.