Open to AI engineering opportunitiesTempe, AZ
Arjun RanjanAI Engineer
I build production agent systems that plan, retrieve, use tools, and ship inside fast, thoughtful products.
MS Computer Science candidate at Arizona State University, working across agent orchestration, semantic retrieval, production AI workflows, and full-stack delivery.
01Operating profile
Designed for the path from prototype to production.
I work where model behavior, reliable systems, and the product surface meet.
Agent orchestration
Tool-aware systems that decompose work, coordinate specialized agents, and keep execution legible.
LangGraph / tool execution / multi-agent planningSemantic retrieval
Memory and retrieval layers designed around task relevance, context continuity, and measurable precision.
Embeddings / semantic memory / deterministic rankingProduction AI
Provider-agnostic workflows with secure tenant boundaries, resilient error paths, and observable outcomes.
AWS Bedrock / PostgreSQL / edge functionsFull-stack delivery
Fast interfaces and pragmatic services built together, with performance treated as a product feature.
Next.js / TypeScript / FastAPI / Spring Boot02Experience
Shipping history, measured in outcomes.
Production AI, full-stack systems, and engineering instruction across fast-moving teams.
Selfmadee.ai
AI Engineer Intern
- 01 / Build vs. buy / $6K saved
Replaced paid e-signature tooling with an in-house DocuSign-style flow for sequential signing, secure links, and audit trails—saving $6K annually.
- 02 / Catalog ops / 04 brands
Turned Amazon and Shopify catalog onboarding into a repeatable AI-assisted mapping flow for four brands.
- 03 / Workflow fabric / 23 × 04
Unified 23 production AI workflows across four providers behind Bedrock, embeddings, and tool execution.
- 04 / Query compression / -90%
Collapsed Gmail count traffic by 90% and replaced 10K-row email transfers with one PostgreSQL response.
- 05 / Tenant defense / 14 + 69
Closed cross-tenant gaps across 14 integrations and hardened error handling for 69 Edge Functions.
Alleo.ai (Techstars ’23)
AI Engineer Intern
- 01 / Agent runtime / 75 → 95%
Re-platformed chat as a tool-orchestrated agent runtime, lifting successful tool calls from 75% to 95%.
- 02 / Memory layer / +60%
Replaced baseline RAG with semantic memory retrieval and raised retrieval accuracy by 60%.
- 03 / Multi-agent / -35% time
Engineered LangGraph workflows for research and planning that cut task completion time by 35%.
- 04 / Performance / 65 → 90
Collapsed duplicated App Router layouts into shared server components and moved mobile Lighthouse from 65 to 90.
Ira A. Fulton Schools — SCAI
Grader (CSE259: Logic in CS)
- 01 / Evaluation system / 120+
Co-designed transparent rubrics that made assessment consistent across a 120+ student cohort.
- 02 / Feedback loop / +15%
Turned grading signals into targeted feedback and office-hour coaching, improving cohort performance by 15%.
Ira A. Fulton Schools — Capstone
Undergraduate Teaching Assistant
- 01 / Engineering mentorship / 70+
Helped 70+ builders turn architecture, testing, and agile trade-offs into software they could confidently ship.
tCognition Inc. (Capstone)
Backend Engineer
- 01 / Secure ATS core / JWT + MongoDB
Designed the authentication and data layer for a high-volume applicant system using Spring Boot, JWT, and MongoDB.
Headstarter
Software Engineering Fellow
- 01 / Ship loop / 03 products
Shipped three production-grade React and Next.js products; CI/CD halved deploy time while backend work cut latency by 40%.
03Selected systems
Two systems. Clear inputs, legible decisions, useful outputs.
Project work framed as architecture and evidence—not a gallery of screenshots.
SYS / 01Case study
AI Flashcards
aranja15 / ai-flashcardsTopic / PDF
FastAPI
Ollama
Card UI
- Problem
- Turn dense topics and uploaded PDFs into usable study material without sending documents through a paid cloud model.
- System
- A React client hands topics and documents to FastAPI, where structured prompts run against local LLaMA inference and parse into card pairs.
- Outcome
- An end-to-end learning flow with file upload, dynamic card state, response parsing, and zero external inference cost.
Stack
React / TypeScript / FastAPI / Ollama
SYS / 02Case study
Agentic Buying Guide
KhushManchanda / clicklessQuery
Planner
Retrieve / Rank
Explain
- Problem
- Translate ambiguous shopping intent into confident purchase decisions across 12K+ headphones and aggregated review data.
- System
- A stateful planner extracts constraints, retrieves and deterministically ranks candidates, then explains each recommendation using supporting signals.
- Outcome
- Reduced irrelevant recommendations by 70% per iteration while preserving budget, feature, and use-case preferences across turns.
Stack
Python / FastAPI / Streamlit / LangGraph / OpenAI
04Technical range
A stack organized around the work.
Model behavior, application systems, and the infrastructure between them.
01
AI systems
LangGraph · LLMs · RAG · Semantic Retrieval · Multi-Agent Systems · Vector Embeddings · Tool Orchestration · AWS Bedrock
02
Application
Next.js · React · TypeScript · Node.js · FastAPI · Python · Java · Spring Boot · REST APIs
03
Data / infrastructure
PostgreSQL · MongoDB · Supabase · Docker · AWS · GCP · CI/CD · Linux · Git
05
Education
Arizona State UniversityDecember 2026
M.S. in Computer Science
Relevant: Semantic Web Mining, Applied Cryptography, KRR
Arizona State UniversityMay 2025
B.S. in Computer Science — 3.92 GPA (Dean’s List, all semesters)
Relevant: DS&A, Compilers, OS, DBMS, ML, Data Mining, iOS, QA, Data Viz