Saurav T.

Key Skills

Software

Mindrift

Snorkel AI

Toloka

Data Annotation Tech

Other

Top Subject Matter

Kubernetes cost analytics and tooling (agent integrations, automation, localization)

Kubernetes security observability and telemetry (eBPF + Prometheus)

AI quiz generation web application (auth/RBAC + performance)

Top Data Types

Computer Code Programming

Text

Top Task Types

Computer Programming/Coding

Evaluation/Rating

Prompt + Response Writing (SFT)

Function Calling

RLHF

Text Summarization

Question Answering

Text Generation

Data Collection

Red Teaming

Fine-tuning

Freelancer Overview

Core Contributor | CNCF Kubernetes Ecosystem | Remote. Brings 2+ years of professional experience across complex professional workflows, research, and quality-focused execution. Core strengths include Other. Education includes Bachelor of Technology, SGGSIE&T, Nanded (2026). AI-training focus includes data types such as Computer Code and Programming and labeling workflows including Function Calling, Computer Programming, and Coding.

Labeling Experience

Augle AI | Software Engineer Intern | Pune, India

Other

Validated live factory software workflows by debugging HMI/IPC interactions and merging production-ready features for PLC-integrated robotic paint automation. Developed automation features including colour mapping, fluid mix ratio, servo configuration, and parameter adjustment. This role centers on industrial automation software rather than dataset creation or AI training data labeling. • Integrated PLC and robotic paint controls • Debugged HMI/IPC workflows for reliable factory operation • Implemented color mapping and mixing/servo parameter controls • Produced production-ready automation feature merges

2026 - Present

AI Code Quality Specialist — Alignerr (Labelbox)

Computer Code ProgrammingComputer Programming/Coding

Onboarded as a Software Developer specializing in AI Code Quality evaluation. The role focuses on assessing AI-generated code outputs for correctness, style adherence, security issues, and alignment with engineering best practices across multiple languages and frameworks. Tasks involve reviewing code generated by large language models, identifying logical errors, suggesting better implementations, and providing structured feedback used to refine model behavior. Completed the platform's Zara AI screening interview covering code review judgment, technical communication, and quality calibration. Stack exposure: JavaScript, TypeScript, Python, system design fundamentals, code review best practices.

2026 - Present

Core Contributor | CNCF Kubernetes Ecosystem | Remote

OtherFunction Calling

Built an MCP server for OpenCost to enable AI agents and automation pipelines to query Kubernetes cost analytics, supporting programmatic retrieval of structured data. Added multi-cloud network cost tracking across AWS, GCP, and Azure to improve the breadth and usefulness of the queried analytics. Extended Kubernetes tooling by adding Hindi and Tamil i18n support for Headlamp to broaden accessibility for multilingual operators. • Implemented JSON-RPC 2.0 interfaces for agent-driven cost queries • Integrated multicloud cost tracking for more complete analytics • Enhanced Headlamp with additional localization resources • Developed features to support Kubernetes operator workflows via tooling rather than manual labeling

2025 - Present

SWE-bench Task Author — AfterQuery (Project Silver)

Computer Code ProgrammingComputer Programming/Coding

Authoring SWE-bench-style coding tasks against private production codebases used to evaluate AI agent performance in realistic software engineering settings. Each task consists of: - A natural-language instruction describing a bug or feature request without revealing the implementation - A machine-verifiable test harness (Jest, pytest, go test) with fail_to_pass and pass_to_pass test contracts - A unified-diff reference solution applied via git apply - A Docker environment built from an approved repo base image - Internal root-cause analysis and test plan documentation Submissions pass through a 5-stage validation pipeline: similarity check against existing tasks, LLM rubric review against 11 quality criteria, Dockerfile static checks and image build, parallel null/oracle test runs, and an easiness probe with a frontier model. Tasks must land in the 1–4 of 10 difficulty sweet spot. Stack exposure: TypeScript, Node.js, Go, Python, Docker, PostgreSQL, Redis, gRPC.

2025 - Present

Brainopoly | Software Development Intern | Bengaluru, India

Other

Engineered a full-stack platform that includes AI quiz generation alongside JWT authentication and role-based access control. Optimized Prisma APIs and RBAC dashboards to maintain sub-200ms query latency on high-traffic endpoints. The AI component described is quiz generation, but no explicit data labeling or annotation workflow is specified. • Built React/TypeScript frontend with Node.js backend services • Implemented JWT auth and RBAC features • Integrated AI quiz generation into the platform • Tuned Prisma APIs and dashboards for low-latency performance

2025 - 2025

Show more

Education

Work History