Skip to content
View Nsujatno's full-sized avatar

Highlights

  • Pro

Block or report Nsujatno

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
Nsujatno/README.md

Nathan Sujatno

AI Engineering Intern @ Valero Energy · CS + Intelligent Systems @ UT Dallas

Portfolio LinkedIn GitHub Email

Richardson, TX


About

CS + Intelligent Systems (fast-track master's) student at UT Dallas, graduating May 2027 (BS) / May 2028 (MS), Hobson Wildenthal Honors College. Currently an AI Engineering Intern at Valero Energy, building an LLM-powered document extraction pipeline and its evaluation harness.

Focused on AI/agentic systems engineering — retrieval, evaluation, and multi-agent orchestration — with a full-stack background (React/Next.js, FastAPI/Node, AWS). More interested in whether a system is reliable and explainable than in how flashy the demo looks.

Open to: Spring 2027 AI engineering internships


Tech Stack

Languages Python TypeScript SQL

AI / ML LangGraph LlamaIndex PyTorch Scikit--learn ClearML

Backend & Data FastAPI Node.js MongoDB Supabase Pinecone

Frontend & Cloud React Next.js Tailwind AWS Azure DevOps


AI / ML Focus Areas

Domain Details
Retrieval & RAG BM25, dense, and hybrid (RRF) retrieval; benchmarking against hand-labeled golden datasets
Evaluation LLM-as-judge scoring (faithfulness, relevance, completeness); automated eval pipelines with ClearML
Agentic workflows Multi-agent orchestration and routing with LangGraph
Document extraction LLM-powered structured extraction from unstructured enterprise data (Docling + LlamaIndex)

Featured Projects

Ask My Doc — RAG evaluation framework (in progress, targeting Sept 2026)

A production-style evaluation harness for RAG over HuggingFace documentation. Benchmarks BM25, dense, and hybrid (RRF) retrieval against a hand-labeled golden dataset, scored with an LLM-as-judge on faithfulness, relevance, and completeness.

Stack Python, pgvector, text-embedding-3-small, Gemini 2.5 Flash, Streamlit, Docker
Focus Retrieval strategy comparison + eval rigor, not just pipeline construction
Status In development
GreenGain — 2nd Place, HackEarth (Feb 2026)

Multi-agent workflow that routes vector retrieval strategy (local vs. hybrid) based on user geographic data to produce location-specific tax recommendations, with a validation layer enforcing hard financial constraints.

Stack Python, LangGraph
Reliability Programmatic filter guarantees 100% compliance with ROI thresholds under 30 years
Result 2nd place, HackEarth
Eco Quest — 1st Place / 1,000+ participants, Dev Season of Code (Feb–Mar 2026)

Personalized carbon-reduction assistant: a LangGraph workflow generates missions from survey data and CO2 calculations, a Gemini Vision pipeline scans receipts to estimate footprint, and an agentic shopping assistant streams personalized suggestions using purchase history from Supabase.

Stack LangGraph, Gemini 2.5 Flash, Gemini Vision, Supabase, SSE streaming
Result 1st place / 1,000+ participants
OPTCG Binder App — One Piece TCG binder builder

Web app for building and organizing One Piece TCG binder pages, with an AI feature that uses embeddings to suggest visually or thematically similar cards.

Stack Embeddings-based similarity search

Experience

AI Engineering Intern — Valero Energy Corporation, San Antonio, TX May 2026 – Current

  • Owning end-to-end development of an LLM-powered document extraction system (Docling + LlamaIndex) reducing manual processing time by ~95%
  • Built an automated ClearML evaluation pipeline benchmarking extraction accuracy at 90% against a ground-truth dataset, removing manual validation bottlenecks

LlamaIndex Docling ClearML LLM Evaluation

Fullstack Intern — NRVE (AI-powered mental health app for kids), Remote June 2025 – August 2025

  • Built a serverless AWS backend (Lambda, API Gateway, S3) + MongoDB with 15+ REST endpoints
  • Cut audio endpoint response time from 3.5s to 40ms using CloudFront edge caching
  • Built the backend-to-mobile integration layer (React Native, Axios) including mini-game state and session management

AWS Lambda CloudFront MongoDB React Native


Leadership

ACM Dev Lead & Officer — MeteorMate June 2025 – Current Co-leading a 7-person team building an AI-powered roommate matching platform for UTD students — sprint planning, task delegation, code review.

Project Manager — AI MD, UTD Student Organization September 2025 – December 2025 Directed a team of 4 developers building a full-stack MERN healthcare application.


Achievements

Recognition Details
1st Place / 1,000+ participants Dev Season of Code Hackathon — Eco Quest
2nd Place HackEarth — GreenGain
Academic Excellence Scholar Hobson Wildenthal Honors College, UT Dallas

Pinned Loading

  1. acmutd/meteormate-client acmutd/meteormate-client Public

    Client website for MeteorMate

    TypeScript

  2. Intellect-Ink Intellect-Ink Public

    Forked from acm-projects/Intellect-Ink

    TypeScript

  3. kanban-sync kanban-sync Public

    TypeScript

  4. swe-job-matcher-fullstack swe-job-matcher-fullstack Public

    TypeScript 1

  5. portfolio portfolio Public

    TypeScript

  6. MCP-server-for-notes MCP-server-for-notes Public

    Python 1