Udit Bhatt

Software Development Engineering Intern

Talent

HaldwaniMember since 1 June 2026

LangChainPrompt EngineeringOpenRouterYOLOv8PythonFastAPINode.js/ExpressJWTMongoDBMotorSQLVercelReactNext.jsMapboxWebXRJavaC++TypeScriptRust

Hire this person

Send a job offer directly to this candidate

LangChainPrompt EngineeringOpenRouterYOLOv8Python

About

Shipped a production FastAPI + React (Vite) platform that ingested gigapixel UAV orthomosaic images end-to-end, replacing a manual review workflow and cutting analysis time from hours to minutes per field report. Engineered a tile-based rasterio pipeline with sliding-window chunking and global Non-Maximum Suppression; a rule-based junk filter reduced false-positive detections (soil, shadow, background) by suppressing noise across the full image before results were surfaced to users. Integrated YOLOv8 plant localization with a three-class health classifier (healthy / mild / severe), backed by MongoDB (Motor) with JWT role-based auth — the complete stack went from local prototype to a deployed, multi-user web service within the internship tenure.

Experience

Software Development Engineering Intern - Mahyco Grow

(2026-01 - 2026-03)

Shipped a production FastAPI + React (Vite) platform that ingested gigapixel UAV orthomosaic images end-to-end, replacing a manual review workflow and cutting analysis time from hours to minutes per field report.
Engineered a tile-based rasterio pipeline with sliding-window chunking and global Non-Maximum Suppression; a rule-based junk filter reduced false-positive detections (soil, shadow, background) by suppressing noise across the full image before results were surfaced to users.
Integrated YOLOv8 plant localization with a three-class health classifier (healthy / mild / severe), backed by MongoDB (Motor) with JWT role-based auth — the complete stack went from local prototype to a deployed, multi-user web service within the internship tenure.

Machine Learning Analyst (Freelance) - Shipd by Datacurve - Remote

(2026-02)

Design private evaluation benchmarks for foundation model labs that test coding-agent performance, multi-step reasoning, and real-world ML engineering skill across Java, C++, TypeScript, Python, and Rust.
Maintain a >90% code-approval rate after automated and manual review across 5 languages — directly informing LLM training signal for production model releases.
Build end-to-end benchmark pipelines and submission workflows adopted by internal and open-source LLM evaluation projects, bridging the gap between research prototypes and reproducible eval infrastructure.