Welcome to the space of
Pranav Lokhande
Final-year Software Engineering & Data Science Student at University of Sydney
Pranav Lokhande
AI / ML Engineering Intern @ Humello · USYD
About Me
I'm a final-year Software Engineering and Data Science student at the University of Sydney with experience building full-stack applications, AI-powered tools, and cloud-based systems.
I'm currently an AI / ML Engineering Intern at Humello, building a RAG system with Qwen3 LLMs, BGE embeddings and pgvector, and deploying models with vLLM on NVIDIA GPUs. Before that I built an AI-driven geospatial platform at the NSW Decarbonisation Hub. I'm comfortable across the stack, from React and Next.js frontends to Java/Spring Boot and Node.js backends, REST APIs and cloud deployment on AWS.
My Honours research looks at non-LLM anomaly detection for multi-agent LLM systems, at the intersection of agentic AI and cybersecurity. Lately I've also been building my finance knowledge through CFI financial modelling coursework.
I spend free time reading about design systems, experimenting with datasets, and refining workflows to ship faster with fewer regressions. When I'm not coding, you'll find me at the beach or playing cricket with friends.
University of Sydney
Bachelor of Engineering (Honours) - Software, Data Science Specialisation
Ambassador
Mentor
Representative
Assistant
Nanyang Technological University
Singapore
Introduction to Data Science, AI & Cybersecurity
Singapore Management University
Singapore
Quantum Computing in Finance Services
Impact that spans research and product.
AI / ML Engineering Intern
Humello
Developing an AI-powered RAG system for Australian Modern Awards using Python, Qwen3 LLMs, BGE embeddings, reranking, and PostgreSQL/pgvector for accurate award and classification retrieval
Built query-routing and AI automation workflows and deployed Qwen3-14B with vLLM on NVIDIA GPUs, integrating APIs, debugging retrieval pipelines, and securing model infrastructure and network access
Honours Thesis Researcher
CSIRO's Data61 & University of Sydney
Researching anomaly detection in multi-agent LLM systems, developing controlled Python environments to detect malicious agent behaviour from communication and interaction data
Developing and testing ML and graph-based detection methods against LLM-as-a-Judge and existing baselines using reproducible experiments and performance metrics
Software Engineering Intern – P2X Internship
NSW Decarbonisation Hub
Engineered an AI-driven geospatial web application using React, Node.js, and PostgreSQL/PostGIS, integrating REST APIs and environmental datasets across 50k+ property listings
Developed LLM-powered natural-language search and optimised full-stack data workflows and interactive mapping for flood, bushfire, air-quality, and property data
Research Assistant
View projectUniversity of Sydney
Developed LLM-powered educational tools leveraging OpenAI to automate course-related queries, providing intelligent responses to student questions and streamlining administrative workflows for teaching staff
Engineered data pipelines for preprocessing, orchestration, and real-time monitoring
Developed Streamlit web application automating LCMS Instruments data generation with Python
Summer Researcher
Charles Perkins Centre
Optimized Python-based data processing pipelines for large-scale analysis
Built Python visualization libraries and automation scripts
Integrated Python tooling into research workflows for efficiency
Projects with narrative and scale.
LoopyPool
Tap to learn moreHover to learn more
LoopyPool
Australia-wide 3D geospatial property intelligence platform built on Google Photorealistic 3D Tiles, with a Node.js API over a 2M+ record PostgreSQL database for viewport-based spatial queries and natural-language property search. Deployed on AWS (EC2, S3, CloudFront).
Cypher · Agentic VC Analyst
Tap to learn moreHover to learn more
Cypher · Agentic VC Analyst
Top 10 hackathon finalist. A 34-agent platform that automates investment workflows from pitch-deck ingestion to valuation and IC recommendation, with founder analysis, market sizing, DCF/VC modelling and bull/bear debates.
Arceus · YouTube Virality
Tap to learn moreHover to learn more
Arceus · YouTube Virality
Full-stack prediction platform trained on 151K+ trending YouTube videos across 11 markets, combining TF-IDF, SVD-200 embeddings and market-aware features to predict cohort percentile rankings.
Predicting Building Energy Consumption
Tap to learn moreHover to learn more
Predicting Building Energy Consumption
Benchmarked CNN, BiLSTM, ConvLSTM and LightGBM for 24-hour-ahead load forecasting across 20 commercial buildings from Building Data Genome 2, with LightGBM outperforming deep learning in 80% of buildings.
MediaTracker (Medley)
Tap to learn moreHover to learn more
MediaTracker (Medley)
A comprehensive full-stack media tracking application for discovering, organizing, and tracking books, movies, TV shows, anime, and manga with advanced search, custom lists, and social features.
Group Allocation Management System
Tap to learn moreHover to learn more
Group Allocation Management System
A full-stack web application for managing student project groups, task allocation, and collaboration in university environments with role-based access, real-time progress tracking, and workload balancing.
Trains, Trends & Turbulence
Tap to learn moreHover to learn more
Trains, Trends & Turbulence
Analyzed Sydney ridership vs. disruptions with geospatial storytelling, statistical overlays, and decision-ready dashboards.
Supply Chain Intelligence
Tap to learn moreHover to learn more
Supply Chain Intelligence
A comprehensive data science project analyzing supply chain logistics data to infer transport modes and understand their relationship with risk metrics using unsupervised learning and predictive modeling.
Signals from mentors and juries.
Charles Perkins Centre Summer Research Scholarship
University of Sydney
Selected among 19 students to drive metabolomics research with automation-first data tooling and rapid iteration.
Winter Data Analysis Challenge · Honourable Mention
Sydney Precision Data Science Centre × Westpac
Analysed Sydney rail ridership from 2020–2025, combining transport, weather and event datasets with spatial and statistical methods.
Cypher · Top 10 Finalist
OpenAI × January Capital × Lyra × Relevance AI Hackathon
Built an agentic VC analyst platform with multiple AI agents for market, competition, founder and valuation analysis.
Certifications
AWS · ISC2 · Microsoft
AWS Certified Cloud Practitioner · AWS Certified AI Practitioner · ISC2 Certified in Cybersecurity (CC) · Microsoft AZ-900 Azure Fundamentals
Let’s build the next niche system.
Drop me a pulse for collaborations, residencies, or research pairings. Dark mode comes first, but I obsess equally about how the light version behaves.
Availability
• Accepting freelance and part-time systems collabs.
• Happy to jam on AI strategy or lightning-fast prototypes.
• Based in Sydney, working async across timezones.