Senior Full Stack Engineer · React · Next.js · NestJS · Generative AI · LLM Apps · Prompt Engineering
Based in Ahmedabad, India · 10+ years building enterprise-grade applications
I build full stack applications at the intersection of modern web engineering and Generative AI. During the day I lead frontend modernisation at LTM (Larsen & Toubro Group), rewriting legacy enterprise systems into scalable React and Next.js applications. In my own time I architect AI-integrated SaaS products from scratch.
I am especially interested in multi-model LLM orchestration, IMG2IMG AI pipelines, and building real products that use Generative AI practically — not just as a demo.
Upload one product photo → get 10 professional studio-quality images in under 3 minutes
- Hybrid AI pipeline: Gemini 3 Pro Image for hero shots + Replicate FLUX.1 Kontext for batch angle generation
- Dynamic angle selection: Gemini Vision analyses each product and recommends the best angles based on garment type, silhouette, and user's photo count setting
- 61% cost reduction achieved through provider-agnostic architecture — from $0.67 to $0.26 per 10-angle job
- LoRA fine-tuning pipeline planned on FLUX.1 Kontext Dev via Replicate for category-specific photoshoot style training
- Stack: React Native (Expo) · NestJS 11 · PostgreSQL · Redis · BullMQ · AWS S3 + CloudFront · Razorpay · Stripe
Models worked with:
├── Google Gemini 3 Pro Image → Hero product shot generation (IMG2IMG)
├── Google Gemini 3.1 Flash Image → High-volume angle generation
├── Google Gemini 2.5 Flash → Vision validation + structured JSON extraction
├── OpenAI gpt-image-2 → Image generation (low/medium/high quality)
├── Replicate FLUX.1 Kontext Dev/Pro → IMG2IMG product photography + LoRA fine-tuning
└── Microsoft Copilot → Agentic AI application development
Techniques:
├── Prompt engineering (image-led, category-specific, minimal token prompts)
├── Gemini thinkingConfig multi-image loop (single-call batch generation)
├── Dynamic LLM-based angle selection via vision reasoning
├── Hybrid multi-provider orchestration with provider-agnostic service architecture
├── Token economics and AI cost modelling ($/image optimisation)
└── LoRA fine-tuning pipeline design for custom model training
| Company | Role | Period |
|---|---|---|
| LTM - Larsen & Toubro Group | Specialist — Industrial IoT iNxt | 2022 – Present |
| SPEC INDIA | Senior Software Developer | 2021 – 2022 |
| KGN Technologies | Senior Software Developer | 2019 – 2021 |
| Red Turtle India | Software Developer | 2018 – 2019 |
| Shiponk | Software Engineer | 2016 – 2018 |
Key deliveries:
- MES (Manufacturing Execution System) — complete .NET XBAP → ReactJS rewrite for Carrier Corporation
- Shop Floor Display (SFD) — real-time production tracking dashboard for Carrier Corporation shop floors
- Impact Aclar Calculator — data-driven packaging material comparison tool for Honeywell (94+ Lighthouse score)
- Instrucko — English learning platform (40% load time improvement via React Profiler optimisation)
- 🏆 AZ-900 Microsoft Certified Azure Fundamentals — Microsoft
- 🤖 Agentic AI and AI Agents: A Primer for Leaders — Vanderbilt University
- 🧠 Generative AI: Working with Large Language Models — LinkedIn Learning
- 💬 Prompt Engineering for ChatGPT — Coursera
- ⚛️ React Certified — Triplebyte
- 🏢 iLead Program — LTM
| Tool | Use case |
|---|---|
| Claude CLI | Architecture decisions, complex prompt engineering, code review |
| OpenAI Codex | Code generation and refactoring |
| Google AI Studio (Antigravity) | Gemini API testing, multi-image generation research |
| Factory.ai | Agentic development workflows |
"I don't just use AI tools — I build the infrastructure that runs them."
