00 / INDEX

CS · UT ARLINGTON

DALLAS, TEXAS

AI/ML + CV

AI Engineering · Full Stack · Machine Learning · Computer Vision

Building AI systems from model to real-world products.

Computer Science student at UT Arlington focused on AI engineering, machine learning, and computer vision. I build end-to-end intelligent systems - from models and real-time vision pipelines to full-stack products people can actually use.

watch me build ↗github.com/Pranillama

Applied AI DeveloperDallas AI

Undergraduate ResearcherOURCS@DFW

Data Analytics InternHunt Consolidated, Inc.

Applied AI DeveloperDallas AI

Undergraduate ResearcherOURCS@DFW

Data Analytics InternHunt Consolidated, Inc.

Applied AI DeveloperDallas AI

Undergraduate ResearcherOURCS@DFW

Data Analytics InternHunt Consolidated, Inc.

Applied AI DeveloperDallas AI

Undergraduate ResearcherOURCS@DFW

Data Analytics InternHunt Consolidated, Inc.

02 / FOCUS

where my work
concentrates

AI_ENGINEERING

sys.01

AI Engineering

Building end-to-end AI-powered applications — integrating models, APIs, retrieval, structured data, and full-stack interfaces into reliable products people can actually use.

ML_SYSTEMS

sys.02

Machine Learning

From neural networks and deep learning to embeddings and retrieval — training, evaluating, and integrating ML models to solve real-world problems.

VISION

sys.03

Computer Vision

Building and deploying real-time computer vision systems — from model training and inference pipelines to local and edge deployment for detection, tracking, and visual understanding.

AGENTS

sys.04

Agentic Systems

Exploring AI agents, tool use, memory, workflow automation, and voice interfaces — with a focus on building systems that can reason, act, and interact beyond a single prompt.

03 / LOG

where i've
worked

Experience

Applied AI DeveloperMay. 2026 – Aug. 2026

Dallas AI · Dallas, TX

watch final demo ↗

Decided to replace a traditional CV pipeline with a multimodal vision-language-model approach under a tight MVP timeline, weighing build-vs-buy tradeoffs to hit the deadline.
Served as Technical Lead for an 8-person team, shaping architecture decisions and breaking down implementation work across developers and PMs with varying technical backgrounds.

Undergraduate ResearcherFeb. 2025 – Mar. 2025

OURCS@DFW · Arlington, TX

view research certificate ↗

Applied adversarial attacks to text summarization models and measured robustness impact using Hugging Face and ROUGE scores, uncovering a 10–15% drop in recall on adversarial samples.
Investigated AI security vulnerabilities and model-evaluation techniques, translating findings into concrete prompt-template recommendations.

Data Analytics InternSept. 2023 – Dec. 2023

Hunt Consolidated, Inc. · Dallas, TX

view internship work ↗

Analyzed qualitative and quantitative datasets to support a $500K+ investment decision, cleaning and structuring data in Excel to cut report-preparation time by 30%.
Synthesized insights from 15+ stakeholder interviews and 5+ site visits using AI-assisted summarization, accelerating technical decision-making and executive-ready reporting.

04 / PROJECTS

Selected work

8 shown

01Multi-Agent

ROLE

AI Systems

second_brain_agent

Gemini · Google ADK · Cloud Run4 AgentsLive

Live AI second brain that watches an Obsidian vault, organizes incoming notes into a structured knowledge base, and generates nightly project briefs.

›Designed a four-agent event-driven pipeline with isolated write boundaries and human approval gates for task and project-status changes.

GeminiGoogle ADKCloud RunFirestoreCloud StorageMulti-Agent · Event Driven · Agentic Workflows

saving_water

TypeScript · PythonHackSMU VII24h Build

Water-usage intelligence platform that turns household consumption data into actionable conservation insight. Winner, Overall 3rd Place at HackSMU VII with a team of three.

›Owned UI/UX and the RAG system, which uses attribute-based matching instead of a vector database.

TypeScriptPythonUI/UX · RAG · Full Stack

devpost ↗

jhakri-ai

Next.js · RealtimeRealtime

Collaborative system-design workspace: describe an architecture in plain English, generate it on a shared canvas, refine it with collaborators, and export a technical spec.

›Designed the LLM-to-diagram generation loop and the realtime multiplayer canvas that keeps concurrent architecture edits consistent across collaborators.

Next.jsReactFlowLiveblocksTrigger.devRealtime · Distributed State · AI Systems

VisionGuide

On-device CV · VoiceOn-DeviceRealtime

Mobile app for real-time indoor navigation assistance — on-device object detection paired with spoken guidance for accessibility.

›Built the on-device perception loop — MediaPipe detection fused with Claude Vision scene reasoning — driving spoken guidance at interactive latency.

MediaPipeClaude VisionWeb Speech APIReactComputer Vision · Realtime AI · Edge AI

github ↗

ROLE

Computer Vision

MavPark

YOLOv8 · Spring BootPrototype

Prototype parking detection for UT Arlington with YOLOv8, designed for edge deployment on a Jetson Nano.

›Worked on the YOLOv8 detection pipeline for the prototype, designed for a Jetson Nano and a Spring Boot API.

YOLOv8PythonJava (Spring Boot)ReactComputer Vision · ML Pipeline · Edge

github ↗

LimenFit

Docker · PostgreSQL

Minimal, high-speed fitness tracker: fast workout logging, AI-assisted progress tracking, plan sharing, and computer-vision form analysis.

›Built the pose-landmark form-analysis layer on top of a containerized PostgreSQL backend designed for sub-second workout logging.

TypeScriptPostgreSQLDockerFull Stack · Computer Vision

github ↗

CloudTask

AWS ServerlessServerless~$0 Idle

Full-stack serverless task manager on AWS with scheduled daily reminders — a study in event-driven, low-cost cloud architecture.

›Architected an event-driven serverless backend — EventBridge schedules fan out through Lambda and SNS — running at near-zero idle cost.

LambdaDynamoDBEventBridgeSNSCognitoServerless · Event Driven · Cloud Infrastructure

github ↗

08RAG

Music Recommender w/ RAG

Gemini · GuardrailsGrounded

Extends a rules-based recommender into a full RAG pipeline: embeds natural-language queries, retrieves matches via similarity search, and generates conversational recommendations.

›Implemented the retrieval pipeline and a grounding guardrail that rejects any recommendation not traceable to retrieved catalog data.

Gemini APIEmbeddingsPythonRAG · ML Pipeline

github ↗

05 / STACK

tools &
capabilities