Hello

I am Sarvesh Patil,
an AI Engineer
& Full Stack
Developer.

About

Master's student in Computer Science at NC State with 2 years of software engineering experience at Nomura. I build AI-powered applications, intelligent automation tools, and scalable software systems, with experience across LLMs, agentic AI, Python, and full-stack development. Currently exploring Generative AI, LLMs, and AI engineering, with a focus on turning emerging AI capabilities into reliable, production-ready solutions.

Download Resume

Expertise

AI Engineering
Software Development
LLM & Agentic AI
Full-Stack Engineering
Data Engineering
System Design

Experience

Cut manual Change Assessment reviews & pre-submission rejections by building a dual-agent Backstage plugin assessing requests against 19 compliance criteria; automated confluence & attachment ingestion, timestamp verification, & output guardrails with Japanese-localized outputs using LLMs, Python & React.js.
Gave senior management visibility into access-control violations across 5 regions & 40 divisions by building an analytics dashboard with Sankey flows, live KPIs, & SSE updates, backed by ETL pipeline using React.js & Python.
Strengthened platform security & reliability by implementing token-based API authentication, ETL retry logic, & session caching that cut API latency by 80%, plus scheduled MongoDB backups enabling recovery within 1 hour.

Eliminated a 30-day manual reporting cycle across 120+ application teams by building a real-time governance dashboard unifying incident, change, infringement & configuration data, using React.js, TypeScript & MUI.
Boosted GitLab CI adoption by 25% by building a unified CI/CD plugin consolidating GitLab CI & Jenkins data from 500+ repositories into one dashboard, enabling data-backed Jenkins deprecation decisions using GraphQL.
Reduced release & incident analysis time from hours to seconds across 500+ applications by building AI-powered ETL pipelines with on-demand force-refresh & prompt chunking for large inputs, delivering automated root-cause analysis & management-level summaries using LLMs, Python, & FastAPI.
Reduced pitchbook creation time from days to hours & standardized output across 100+ annual pitchbooks by leading a 6-engineer team to build an AI automation platform with document assembly, chart/map generation, & natural-language editing using LangChain Agents, FastAPI, Next.js, Tailwind CSS, Matplotlib, & Geopy.

Eliminated paper-based workflows for 1,500+ sanitation facilities across 25 wards by digitalizing toilet community data collection through full-stack web portal & admin dashboard using Next.js, Node.js, Express.js & MongoDB.
Cut server storage costs by 50%, reducing operational expenses, by independently designing & implementing an image compression pipeline using Sharp for facility data uploads.
Delivered a secure, reliable portal across 25 wards by integrating 2Factor OTP auth & automated reporting with jsPDF & XLSX, then deploying & maintaining the application on Amazon EC2 owning the full build end-to-end.

Delivered seamless 3D model search and purchase by integrating Sketchfab and Poly-Pizza APIs with a Razorpay payment gateway, improving user workflow and platform usability.
Strengthened authentication security by implementing Passport.js and bcrypt-based encryption in MongoDB, enhancing data protection and reducing security risks.
Improved user workflow efficiency by architecting the Mongoose schema and building the Login, Register, and Courses pages in React.js/TailwindCSS, reducing latency and enabling smoother interactions.
Stack: React.js, Node.js, Express.js, MongoDB, Passport.js, Razorpay, Tailwind CSS

Improved policy accuracy by implementing validation logic, reducing manual errors.
Enhanced release stability by testing and managing/merging code, cutting deployment issues.
Enabled efficient data sharing by adding PDF dashboard downloads, boosting reporting efficiency.
Achieved clearer product direction by designing Figma wireframes and prototypes, reducing development rework.
Stack: Python, Flask, Figma, Git