Himanshu Kala | AI Quality Assurance Engineer

01 // about_me

About Me

I'm an AI-focused Quality Assurance Engineer with over 2 years of experience in manual and automation testing, with deep specialization in GPT-style LLM evaluation and AI safety validation.

My work sits at the intersection of AI quality and safety — I help ensure that conversational AI systems behave accurately, ethically, and reliably before they reach real users.

From hallucination detection to adversarial prompt design, I test the edges of AI systems so they're ready for the real world. I thrive in Agile environments and collaborate closely with cross-functional teams.

Phone+91 7417517798

Email

LocationMeerut, Uttar Pradesh, India

LinkedInhimanshu-kala-436751235

GitHubHkala402

// profile

RoleAI QA Engineer

Experience2+ Years

CompanySimublade Technology

LocationMeerut, India

Status● Open to Work

SpecialtyLLM Evaluation

02 // skills

Skills & Expertise

AI & LLM Testing

GPT-style LLM EvaluationHallucination Detection AI Safety EvaluationHarmful Content Detection LLM ValidationConversational AI Testing Prompt TestingManual Response Validation

Testing Methodologies

Manual TestingEnd-to-End Testing Smoke & Sanity TestingRegression Testing Test Case DesignDefect Lifecycle Agile / ScrumModel Regression Testing

Tools & Platforms

PostmanJIRA LogRocketRelay RetoolAudacity SlackSQL

03 // work_experience

Work Experience

Oct 2023 – Present ·

Simublade Technology (via Tavus.io)

AI Products · AI Video Personalization · Meerut, India

Associate QA Engineer

AI & LLM Testing

Conduct GPT-style LLM evaluation by validating 300+ model responses weekly for contextual relevance, factual accuracy, and logical consistency.
Perform hallucination detection to identify fabricated and misleading AI outputs, reducing model response defects during regression cycles.
Execute AI safety testing to detect unsafe, harmful, biased, or policy-violating responses.
Design structured, adversarial, and edge-case prompts to test robustness of conversational AI systems.
Conduct regression testing after model updates to ensure stability and performance consistency.

Video & Avatar QA

Validated AI-powered video transformation pipelines ensuring lip-sync synchronization, facial alignment, and voice accuracy.
Re-test conversational AI and digital avatar features after every update to make sure nothing is broken.
Check that the platform stays fast and stable after new AI model releases.

Voice Training & API Testing

Use Audacity to record, clean, and prepare voice audio samples used for training AI speech models.
Test REST APIs using Postman to verify that requests, responses, authentication, and error messages all work correctly.

General

Work closely with developers, designers, and product managers in an Agile/Scrum team to ship AI features on time.
Log, track, and manage defects across full defect lifecycle using Relay and LogRocket.

04 // education

Education

2022

Bachelor of Computer Applications (BCA)

Chaudhary Charan Singh University

Computer Applications (General)

2019

12th Standard

Uttar Pradesh Board

English Medium

2016

10th Standard

CBSE Board

English Medium

✓ Certified

Aug 2022 – Mar 2023

Java Expert (JAVAEXPERT)

DUCAT India

Cert No: 31122022884189904 · Student ID: 31236/2022

05 // projects

Projects

BUILT PROJECT · AI TOOL

AI Test Case Generator

A web-based AI tool that generates structured QA test cases (Positive, Negative, Edge Case) from a plain-English feature description — built end-to-end and deployed live. Demonstrates prompt engineering, LLM output validation, and QA-focused product thinking.

Designed structured prompts that reliably return JSON-formatted test cases from Google Gemini, with defensive parsing and error handling.
Built with vanilla HTML, CSS, and JavaScript — no frameworks, no build step — matching the dark cyan terminal aesthetic of this portfolio.
Configurable test count (3–15), test types, and three output formats: Formatted Cards, JSON, and Markdown Table.
One-click export to JSON or CSV for direct use in test management tools.

Live Demo View Code

BUILT PROJECT · AI DIAGNOSTIC TOOL

Phoenix-4 Replica Rejection Analyzer

An AI-powered diagnostic tool that analyzes why a Tavus Phoenix-4 training video gets rejected — before the customer even submits it. Validates 7 official Phoenix-4 requirements and uses Gemini AI to write a plain-English fix plan with approval probability estimate.

Built a 7-point automated checklist engine validating consent statement, codec, file size, silence segment, lip movement, and video source — with pass / warn / fail logic.
Integrated Google Gemini API to generate expert-level rejection diagnosis and a numbered step-by-step fix plan tailored to each submission.
Designed a 4-step wizard UI with animated score ring, approval probability meter, and terminal-style AI output — built in vanilla HTML, CSS, and JavaScript with zero dependencies.
Includes offline fallback mode — full checklist results shown even without an API key.

Live Demo View Code

WORK PROJECT

AI Video Automation — Tavus AI Integration

Quality assurance for an AI-powered video platform that turns simple inputs into fully personalized, AI-generated videos.

Checked that the AI correctly turned user inputs into personalized videos without errors or quality issues.
Reviewed Phoenix AI model results to ensure faces looked natural and voice matched lip movements properly.
Re-tested AI chat and digital avatar features after each update to confirm nothing was broken.
Monitored the system after new releases to verify it stayed stable under real conditions.

CONVERSATIONAL AI

Digital Human Interaction Testing

Comprehensive validation of digital human and conversational AI interaction systems for production reliability.

Performed end-to-end testing of conversational AI workflows.
Validated digital avatar lip-sync and voice accuracy pipelines.
Ensured high response accuracy standards before production deployment.

06 // achievements

Achievements

⚡

Efficiency

Improved AI Response Validation

Improved AI response validation efficiency by implementing a structured prompt testing strategy across teams.

🛡

AI Safety

Strengthened Safety Compliance

Strengthened AI safety compliance through systematic harmful-content detection workflows.

🔍

Recognition

Recognized for Debugging Skills

Recognized for analytical debugging skills and proactive communication during AI product releases.

🎯

Quality

100% Pre-Deployment Accuracy

Consistently ensured high model response accuracy and safety before every production deployment.

07 // contact

Contact Me

I'm open to new opportunities, collaborations, and conversations about AI quality and safety. Feel free to reach out any time.

Email hkala402@gmail.com

Phone +91 7417517798

LinkedIn himanshu-kala-436751235

GitHub Hkala402

Location Meerut, Uttar Pradesh, India