Headline

Generative AI Engineer | RAG · Agentic AI · LLM Systems | Built infografix.app (15K+ MAU) | Python · LangGraph · FastAPI · Azure OpenAI | Open to AI Roles

About

I build AI systems designed to perform in production — not just in demos. At Infografix (infografix.app), I led AI engineering for a GPT-4-powered infographic platform, helping scale it from an early concept to 15,000+ monthly active users — work that earned me two promotions in under three years (Software Engineer → Software Specialist). I owned the AI stack end-to-end: retrieval pipelines, prompt versioning, schema-enforced structured outputs, evaluation workflows, multi-model routing, and AWS infrastructure. My focus is the unglamorous part that makes Gen AI actually work at scale: latency, inference cost, output reliability, observability, and graceful failure handling. What I work with: • RAG pipeline design — LlamaIndex, LangChain, ChromaDB, FAISS, Pinecone • Multi-agent orchestration — LangGraph, ReAct, function/tool calling • LLM evaluation — LangSmith, RAGAS, hallucination detection • Models — GPT-4/4o, Claude, Llama 3, Mistral • Full-stack AI apps — FastAPI, NestJS, Angular, Docker, AWS Recent projects: • Maestro — an autonomous multi-agent system that hit 85%+ task completion on a custom reasoning benchmark. • Sentinel — a hallucination-detection system (DeBERTa + ChromaDB) that cut ungrounded outputs by 62%. I'm now open to Senior Generative AI Engineer roles where I can design and scale AI products real users depend on. 📬 Reach me: mehtasparsh@outlook.com

Location

Bengaluru

AI stack

claude

cursor

langchain

langgraph

langsmith

openai

Work experience

Software Specialist at Quark Software Inc., Jul 2025 - Jul 2026 · 1 y, On-site, Lead AI-feature delivery on Infografix (infografix.app) — a canvas-based, GPT-4-powered infographic generator. • Own the AI content-generation pipeline end-to-end: Azure OpenAI / GPT-4 integration, per-template prompt design, and JSON-schema-enforced structured outputs across 12+ infographic designs. • Built the plugin content-generation endpoint (NestJS) powering persistent, template-specific AI content. • Drive prompt quality across 12+ designs (Family Tree, Mindmap, SWOT, Timeline, charts, Q&A, Hierarchy, Process, Cycle) with a continuous prompt-iteration cadence and schema-validated outputs. • Serve as sole frontend architect and mentor engineers on AI-integration patterns and production observability. • Impact: contributed to scaling the platform to 15,000+ monthly active users.

Software Engineer 2 at Quark Software Inc., Apr 2025 - Jul 2025 · 4 mos, Chandigarh, India, Expanded into backend ownership and full-stack modernization. Promoted to Software Specialist in ~3 months. • Led the NestJS 10 → 11 migration and Node 18 → 20 upgrade; resolved all flagged Snyk & SAST security vulnerabilities. • Built the AWS S3 + CloudFront asset pipeline: chunked upload & combine, CDN-served shareable links, and multi-format export (SVG / PNG / JPG / PDF). • Extended the multi-tenant layer: per-tenant API-key lifecycle, rate limiting

Software Engineer 1 at Quark Software Inc., Aug 2022 - Apr 2025 · 2 yrs 9 mos, Chandigarh, India, Joined as the first dedicated frontend engineer on Infografix and grew into sole frontend architect across 15+ releases (v1.0 → v2.17+). • Delivered the Angular 16 → 19 migration — the largest modernization in the product's history — with zero downtime and full feature preservation. • Built the custom editor stack (editor, table editors, font picker, theme dialog) and a queue-based undo/redo system with template-switch isolation and typing-sync protection. • Engineered 10+ interactive templates on JointJS Plus with live data-sync between editor and canvas.

Software Engineer at NovoInvent Software, May 2021 - May 2022 · 1 yr 1 mo, Noida, Uttar Pradesh, India, • Built core UI features for Vocabimate, an English-learning app with 100K+ downloads on Google Play and the App Store, using React Native with a scalable cross-platform architecture. • Integrated REST APIs, Redux state management, and asset caching — reducing average screen load time by 40%. • Shipped feature releases in Agile sprints from a single shared codebase serving both iOS and Android.

Research And Development Intern at Econstruct Design and Build Pvt Ltd, Jun 2020 - Aug 2020 · 3 mos, Bangalore Urban, Karnataka, India

Education

B.E, Computer Science, Chitkara University