I build AI agents and LLM products that ship to real users, not just demos.
Production-grade chatbots, RAG systems, AI agents, and applied ML — from model selection and pipeline architecture to the interface someone actually trusts and uses daily.
AI product engineer first, freelancer second.
I'm Sraj, a full-stack engineer working across AI/ML, GenAI, and applied LLM systems. Most of my experience comes from building real AI products end to end rather than one-off client tickets — model selection, pipeline architecture, backend, and the interface a non-technical user has to trust.
Day to day, that means RAG pipelines and AI agents built on Anthropic and OpenAI APIs, Python and Flask/FastAPI services powering those systems, vision and voice AI features, and React Native apps that put AI in front of real users — all deployed on Docker and MySQL infrastructure I also maintain.
If your project needs someone who can own the whole AI stack — from picking the right model and retrieval strategy to the screen someone opens every morning — that's the work I take on.
AI & ML services, priced by scope — not by hour of guesswork.
Every project starts as a fixed-scope quote. Ranges below reflect typical complexity; a call narrows it to an exact number.
AI & LLM Development
- AI chatbot / assistant₹30K–2L
- RAG chatbot over documents₹40K–2.5L
- AI agent development₹50K–5L
- Vision AI system₹40K–3L
- Voice AI assistant₹60K–3.5L
- Local LLM deployment₹20K–80K
GIS & Satellite Intelligence
- NDVI / crop health system₹25K–1.5L
- Sentinel API integration₹20K–1.2L
- Farm / field dashboard₹80K–5L
- Polygon & raster processing₹15K–1.2L
- Full GIS web application₹1L–8L
Backend & APIs
- REST API build₹10K–80K
- Flask backend₹20K–1.5L
- FastAPI backend₹25K–2L
- Complete backend system₹75K–4L
- Third-party / payment integration₹5K–60K
Machine Learning
- Classification model₹20K–1L
- Prediction / forecasting model₹20K–1.5L
- Computer vision system₹60K–4L
- NLP / text intelligence₹50K–3L
Mobile & Web
- React Native app₹50K–3L
- Expo app₹40K–2L
- Admin dashboard₹30K–2L
- SaaS website build₹1L–8L
DevOps & Data
- Docker + VPS deployment₹10K–50K
- MySQL design / optimization₹10K–80K
- CI/CD pipeline₹20K–80K
- Analytics dashboard₹20K–2L
Selected AI systems in production.
Case studies from applied AI, GenAI tooling, and the products they power — architecture and outcomes, not just screenshots.
Core contributor across the full AgriDoot ecosystem — a mobile app, partner console, CRM, and admin dashboard used to run paid farmer programs and FPO partnerships, including with ITC's ITCMAARS initiative and MBC FPO Organization.
Built a seven-module React Native mobile app, a partner-facing web console, farmer data sync tooling into the CRM, and an admin dashboard with a D3.js state-level map of India, server-side pagination, and GeoJSON-driven filtering for advanced farmer segmentation.
A vessel-tracking dashboard built on Sentinel-1 SAR imagery — detecting and visualizing maritime activity from radar data that works day or night and through cloud cover, where optical satellite imagery fails.
Shipped as a self-contained, single-file HTML/CSS/JS build: no backend dependency for the visualization layer, making it simple to deploy and demo directly from a browser.
A four-step AI wizard that turns a plain-language hardware idea into a complete build package — generating code, documentation, and SVG diagrams through the Anthropic API, driven entirely by prompt design rather than templates.
Started as a React build, then rearchitected in vanilla JS to simplify deployment and cut latency between generation steps.
Development and debugging across Gyan AI's product line, including the native Android app and a kiosk-mode progressive web app built for unattended, public-facing deployment.
Work spanned chat interaction debugging, app stability, and the product documentation used to brief partners on the offering.
Industries I build for.
Technical stack.
The tools I reach for most, grouped by where they earn their place.
AI / ML
- Anthropic & OpenAI APIs
- RAG pipelines
- Ollama / local LLMs
- Computer vision
Backend
- Python
- Flask
- FastAPI
- REST API design
GenAI Tooling
- Prompt engineering
- Structured generation
- Vision & voice AI
- Agent orchestration
Mobile & Web
- React Native
- Expo
- JavaScript / HTML / CSS
Data & Geospatial
- Vector & embedding stores
- Sentinel imagery pipelines
- D3.js mapping
Infrastructure
- Docker
- MySQL
- Nginx / VPS
- CI/CD
How a project runs.
Six stages, same order every time, so you always know what's next.
Discovery
Scope the problem, constraints, and success metric before any architecture decision is made.
Design
System architecture, data flow, and interface direction — reviewed with you before a line of code ships.
Development
Built in reviewable increments, not a single opaque delivery at the end.
Testing
Functional and edge-case testing against real data, not just the happy path.
Deployment
Shipped to production infrastructure — Docker, VPS, or your existing environment.
Support
A defined handover, with optional retainer coverage if the system needs ongoing care.
Engagement & pricing.
Fixed-scope packages for defined projects, or a retainer if you need ongoing capacity.
- One core AI feature or integration
- Fixed scope, fixed price
- Async updates
- 2-week typical turnaround
- Multi-module build (model + backend + interface)
- Architecture review included
- Weekly check-ins
- 30 days post-launch support
- Dedicated monthly hours
- Architecture & AI strategy input
- Priority response
- Infrastructure maintenance included
What it's like to work together.
Feedback themes from recent engagements.
"Handed over a rough problem statement and got back a system that matched exactly how our field team actually works — not a generic dashboard."
"The GIS work was the differentiator. Most freelancers can build a dashboard; very few can actually process satellite imagery correctly."
"Clear communication at every stage, and the AI feature just worked in production — no scrambling after handover."
Frequently asked.
How does pricing actually get decided?
Every project starts with a short discovery call. I scope the exact deliverables, then send a fixed quote within the relevant range — no open-ended hourly billing on defined work.
Do you work with teams that don't have technical staff?
Yes — a large part of past work has been translating a non-technical stakeholder's goal directly into a working system, with plain-language updates along the way.
Can you take over an existing codebase?
Yes. I'll do a short architecture review first so any quote reflects the real state of the code, not assumptions.
What's your typical turnaround?
Starter-scope work is typically 1–2 weeks; full systems run 4–10 weeks depending on integration complexity. Timelines are set during discovery, not guessed upfront.
Do you offer ongoing maintenance after launch?
Yes, either as a fixed post-launch support window included in Professional-tier projects, or as a monthly retainer for continued development.
Notes & writing.
Occasional write-ups on AI systems, LLM engineering, and applied ML.
Why SAR imagery beats optical for continuous monitoring
Cloud cover kills optical satellite pipelines. Here's how Sentinel-1 radar data changes the reliability math.
RAG chatbots: what actually breaks in production
Chunking strategy, retrieval quality, and the failure modes that don't show up in a demo.
Structured generation: getting reliable JSON out of an LLM
Prompt patterns and validation layers that stop a model from silently drifting off-schema.
Have a system that needs building?
Tell me what you're trying to solve. If it's a fit, I'll follow up with scope and pricing within a day.