C++ for problem solving - 2 

Problem Solving (Basic) Certificate

Programming Using Python (SWAYAM Certificate)

C++ 

Rajasthan Police Hackathon 1.0 (Certificate)

C++ for problem solving - 1 

Python (Basic Test) Certificate

Bachelor of Technology - BTech

• Developed interactive web application using Astro.js, JavaScript, TypeScript, HTML, and CSS, implementing SVG-based India Map visualization with optimized performance and responsive design, improving user engagement and load times.
• Built RESTful APIs using Hono.js, TypeScript, and MongoDB with CRUD operations, authentication
middleware, OpenAPI documentation, and Git/GitHub collaboration.
• Website link: https://abcdkbd.com

SDE intern

• Led the end-to-end development of multiple client websites from gathering requirements and prototyping to deploying responsive, production-ready applications working independently as the sole developer.
• Built couponyog.com, a static coupon showcase platform using React.js and Tailwind CSS. Implemented debounced search,
dynamic category pages, and reusable components to display curated coupons for e-commerce and education platforms.
• Developed uspstracking.us, a real-time package tracking tool powered by the TrackingMore API via a custom proxy.
Built a categorized FAQ system with 30,000+ entries using the Gemini API to automate responses to tracking
queries.

Summer Intern

What I build in Dsider: • Engineered an LLM-powered document extraction agent (Python, FastAPI, LangGraph) using a multi-agent "deep agent" architecture that classifies PDF pages, runs 10+ specialized extraction skills in parallel, merges results, and validates output into a structured intermediate JSON schema — automating manual data entry for complex engineering documents across multiple vendor formats. • Integrated AWS Bedrock (Claude) with a tiered model-routing strategy (high/mid/low capability) and confidence-based escalation, plus per-run cost, depth, and wall-time guardrails to control LLM spend while preserving extraction accuracy. • Exposed the agent via FastAPI with S3-backed uploads, JWT/Cognito auth and tenant isolation, async MongoDB (Motor) persistence, and Server-Sent Events for real-time extraction progress. • Built and integrated a table-extraction skill and intermediate-JSON pipeline, validated end to end against real-world engineering documents.

At Dsider, I work on the backend of a simulation platform — mostly Python, FastAPI, AWS, and MongoDB. But the more interesting part of my work has been building the AI pipeline. I built an LLM-powered agent that reads process engineering documents — PDFs generated from tools like Aspen, Syzygy, and CORYS — and extracts structured data for techno-economic simulations. It runs multiple extraction skills in parallel, routes tasks between Claude models based on confidence, and has guardrails around cost and execution time per run. The goal was to eliminate manual data entry as much as possible — and it mostly does. More recently, I've been working on a natural-language-to-BioSTEAM code interface — basically letting subject matter experts describe what they want in plain English and having the system turn it into runnable simulation code. I've also worked on the less visible side of engineering: fixing cross-tenant IDOR vulnerabilities in pathway generation and spec preview, implementing

Software Engineer

Competitive programmer | Full-stack web developer | Knight (Top 2.8%) on LeetCode

leetcode

github

Know about Suresh Chelani's work experience, education, and much more. Suresh Chelani