CHINMAY ARVIND
Résumé
AI / ML ENGINEER

Chinmay Arvind

I build AI agents, machine learning systems, and reliable software infrastructure.

Open to Applied AI, ML, and backend roles.

View projects Read my résumé
chinmay / terminal
$portfolio

Browse my projects or view my résumé.

Run a command
Local portfolio commands18 projects

SELECT A NODE TO EXPLORE

MS COMPUTER SCIENCE · UNIVERSITY OF MINNESOTAUBC COMPUTER SCIENCE HONORS
01 / ABOUTEDUCATION & BACKGROUND

About me

I’m Chinmay, an MS Computer Science student at the University of Minnesota and a UBC Computer Science Honors graduate.

I work on agentic systems, machine learning, retrieval, model evaluation, and software infrastructure.

Expected May 2027

University of Minnesota - Twin Cities

Master of Science in Computer Science

Minneapolis, MN
Selected coursework

Computational Aspects of Matrix Theory · Deep Learning · Machine Learning Fundamentals · Natural Language Processing

May 2025

The University of British Columbia

Bachelor of Science in Computer Science (Honors)

BC, Canada
Selected coursework

Algorithms · Data Structures · Software Engineering · Databases · Parallel Computing · Machine Architecture · Machine Learning · Deep Learning · Artificial Intelligence · Linear Algebra · Probability · Regression Analysis

UPDATES
Sep 2026DynamicMem submitted for reviewSep 2026AI Engineer Internship at VidasmiApr 2026Origin House Hackathon finalistJan 20262nd Place · Analytics for GoodMay 2025Graduated UBC Computer Science Honors
02 / EXPERIENCEEXPERIENCE / 2024–2026

Experience

01
May 2026 – Sep 2026 · Remote

Vidasmi

AI Engineer Intern

Built an autonomous drone mission platform that connects natural-language planning, deterministic safety controls, and flight execution.

  • Agents
  • Autonomous systems
  • Planning
  • Durable workflows
Role details
200+fault-injected flights
98%workflow success
80%planning-time reduction
02
Sep 2025 – May 2026 · Minneapolis, MN

University of Minnesota

Graduate Research Assistant

Built multimodal financial-fraud research across document, transaction, and graph signals, with bounded investigation workflows and reproducible evaluation.

  • Multimodal ML
  • Graph ML
  • Evaluation
  • MLOps
  • HPC
Role details
10+ GBproduction datasets
0.871AUROC
03
May 2024 – May 2025 · BC, Canada

University of British Columbia

ML & Backend Engineer

Built an AI-powered platform that converts PDF and DOCX content into context-aware questions and document-grounded answers.

  • LLMs
  • RAG
  • Document AI
  • Backend
  • MLOps
Role details
15+ GBeducational content
1,000+questions per hour
04
Sep 2024 – Apr 2025 · BC, Canada

University of British Columbia

Teaching Assistant

Mentored capstone software engineering teams and led introductory programming labs.

  • Software engineering
  • Architecture
  • Java
  • Mentorship
Role details
120+capstone students
20+capstone teams
03 / PROJECTSALL REPOSITORIES

Selected projects

Applied AI, model evaluation, and software engineering across backend and ML infrastructure.

Agentic systems/01

Agentic Kubernetes Payment-Incident Investigation & Remediation System

  • Built an agentic incident-response system correlating Kubernetes health, logs, traces, deployment changes, and payment telemetry across reproducible failure scenarios, achieving 91.7% root-cause Recall@1 and Recall@3 in predicate-assisted development replay.
  • Added structured investigation traces, incident memory, deterministic approval gates, GraphQL, Slack notifications, and durable Pub/Sub processing, reducing median investigation time from 11.8 min to 2.9 min.
  • Python
  • LangGraph
  • Kubernetes
GitHub Details
RAG & evaluation/02

Permission-Aware Commercial Lending RAG & Underwriting Evaluation Platform

  • Built a permission-aware lending RAG system across 3,800+ pages of synthetic policy and borrower documents using hybrid retrieval, semantic chunking, selective query rewriting, reranking, and page-level citation enforcement, achieving 86.54% Recall@10.
  • Created an adversarial evaluation set spanning policy interpretation, borrower evidence, missing documents, contradictions, restricted data, and underwriting exceptions, achieving a 92.5% grounded-answer pass rate while reducing unsupported claims to 1.7%.
  • Python
  • FastAPI
  • LlamaIndex
GitHub Demo Details
Model reliability/03

Risk-Adjusted Hallucination Detection

  • Built a calibrated hallucination-risk detector using 4 uncertainty/evidence signals: token uncertainty, self-consistency disagreement, semantic entropy, and groundedness consistency.
  • Improved PHANTOM selective accuracy from 68.3% to 78.6% by abstaining on high-risk answers while preserving 81.9% coverage.
  • Python
  • Qwen
  • Calibration
GitHub Details
ML infrastructure/04

Distributed GPU Inference Serving & Automated Model Release Platform

  • Built a distributed text and multimodal inference platform with GPU-aware routing, continuous batching, prefix caching, and speculative-decoding evaluation.
  • Automated model evaluation, deployment, and rollback orchestration with CI performance gates, an 18-task Airflow pipeline, and MLflow evidence tracking.
34.6 req/ssustained throughput
247 → 116 msmedian client TTFT · separate quantized 3B release benchmark
29% → 58%average GPU utilization
  • PyTorch
  • Ray Serve
  • vLLM
GitHub Details
Backend systems/05

High-Throughput Payment Webhook Reliability & Risk-Scoring Gateway

  • Produced zero duplicate risk-decision records across 1M+ replayed webhook deliveries using shared event IDs and atomic decision publication, with concurrent-retry and worker-restart verification.
  • Achieved 0.95 F1 versus a 0.87 rules baseline on held-out transactions using a calibrated lightweight classifier and reproducible evaluation pipeline.
12,249events/sec · local benchmark
46.46 msp99 acknowledgement
  • Go
  • Rust
  • Redis Streams
GitHub Details
Agentic systems/06

Agentic Quantitative Research Orchestration & Backtesting Platform

  • Built an autonomous quantitative-research system that turned natural-language investment ideas into relevant literature, structured hypotheses, hybrid factor strategies, sandboxed backtests, and iterative research decisions across 30,000+ historical data records.
  • Recorded successful and failed experiments as reusable research memory and training trajectories, reducing average research runtime from 46 min to 18 min.
  • Python
  • LangGraph
  • Polars
GitHub Demo Details
Applied AI/07

Pangea Chat

  • Returns minimal sentence corrections, structured explanations in the learner’s native language, and a CEFR difficulty label.
  • FastAPI validates the request and checks a bounded application cache before calling a schema-constrained language model, with deterministic post-processing to reconcile the corrected sentence, error list, and correctness flag.
LinkedIn recommendation
  • Python
  • FastAPI
  • Pydantic
GitHub Details
04 / RESEARCHRESEARCH / EVALUATION

Research

Jan – Sep 2026

DynamicMem

  • Co-authored DynamicMem, a long-horizon memory benchmark with synthetic user histories spanning 15 months across 16 applications, where agents infer changing attributes, habits, and preferences from scattered activity.
  • Contributed evaluation methodology and failure analysis. The benchmark tests profile reconstruction and personalized assistance at five quarterly checkpoints as each user’s history grows.
  • The study compared five memory systems and traced over 93% of failures to retrieved evidence, revealing difficulty retaining stable facts while replacing information that has changed.
Read paper
  • LLM evaluation
  • Long-horizon memory
  • Model reliability
Undergraduate honors researchSep 2024 – Apr 2025

Undergraduate Honors Thesis

  • Engineered a modular client-server Java-based RL system capable of processing 10,000+ training episodes per session with efficient Q-table updates, enabling students to conduct real-time experimentation of RL strategies.
  • Implemented an epsilon-greedy Q-learning policy, increasing final cumulative rewards of the RL agent by 15% within the first 100 training episodes.
  • Built a Python-Java bridge via Py4J, enabling 15+ teams to use Python ML libraries in a Java game, cutting development time by 50%.
PNRL / Puddle WorldGame of Amazons / Python–Java
  • Q-learning
  • Java
  • Python
  • Py4J
  • SmartFoxServer
05 / ACHIEVEMENTSHACKATHONS & MILESTONES

Achievements

2nd PlaceJan 2026

Analytics for Good Hackathon

Green ROI · corporate travel sustainability

An end-to-end analytics prototype translating technical findings into actionable recommendations for non-technical stakeholders.

  • 24 hrchallenge
Project details
FinalistApr 2026

Origin House Hackathon

Careflow.ai · home-health referral automation

Built a multi-agent referral workflow from discharge-document intake through eligibility, placement, and scheduling.

  • 24 hrhackathon build
  • <1 minintake → routed · hackathon demo
Project details
View project
06 / MORE PROJECTS

More projects

Computer vision

CLIP Semantic Image Search & Retrieval Evaluation

  • Built text-to-image and image-to-image search with CLIP ViT-B/32, a Gradio interface, and cached 512-dimensional embeddings; encoded the 29-photo collection in 0.609 seconds on CUDA.
  • In a 21-query leave-one-out evaluation against the full collection, achieved 95.2% same-category top-1 retrieval (20/21) versus 33.3% (7/21) for the HSV color-histogram baseline—a 61.9-percentage-point difference.
  • Achieved 0.799 mean average precision versus 0.343 for the HSV baseline, while analyzing text-query failures involving exact object counts and ambiguous descriptions.
  • Python
  • CLIP
  • PyTorch
Details
Computer vision

Jersey Number Recognition Boosted

  • Built an end-to-end computer vision pipeline to extract jersey numbers from sports video, combining person tracking, pose-guided ROI cropping, and OCR/scene-text recognition to produce per-player number predictions.
  • Implemented pose-detection pipeline improvements by adding temporal smoothing for keypoints and enabled GPU reduced-precision inference to improve stability and performance.
  • PyTorch
  • OpenCV
  • FlashAttention
GitHub Details
Multimodal ML

EEG Sleep Deprivation Classification

  • Built an end-to-end EEG deep learning framework on 8,300 resting-state EEG epochs to classify normal sleep vs. sleep deprivation and predict session-level SSS/KSS sleepiness scores.
  • Demonstrated transfer of learned EEG representations to downstream sleepiness prediction, achieving 0.692 Pearson and 0.738 Spearman correlation on held-out subjects.
  • PyTorch
  • EEG
  • CNNs
GitHub Details
Applied AI

Drift.ai

  • Built a full-stack AI personal finance platform that replaced a fragmented 7-step manual analysis workflow with a unified product flow for automated drift detection, what-if simulation, and PDF/email report delivery.
  • Engineered a hybrid deterministic + LLM pipeline combining inflation-adjusted spending analysis with local Qwen to run multi-turn behavioral interviews and generate personalized 30-day recovery plans.
  • Next.js
  • FastAPI
  • Supabase
GitHub Details
Browser automation

CleanBreak

  • Built an autonomous subscription-cancellation platform that reduced average cancellation time from 8 minutes to 45 seconds, a 91% reduction, with guarded execution and independent billing verification.
  • Helped 10 pilot users avoid $2,400 in annual subscription renewals through independently verified cancellation outcomes with durable receipts.
  • TypeScript
  • Next.js
  • Playwright
GitHub Details
Model reliability

Disagreement-Aware RAG

  • Built a disagreement-aware RAG system that abstains when predicted disagreement risk is high.
  • Engineered a full-stack app that plots coverage vs. hallucination, achieving ~97% coverage at 3–4% hallucination.
  • FastAPI
  • LlamaIndex
  • Next.js
GitHub Details
07 / TECHNOLOGIES

Tech stack

Languages
  • Python
  • TypeScript
  • Java
  • C
  • R
  • JavaScript
  • SQL
  • HTML
  • CSS
LLMs, agents & retrieval
Machine learning
Backend & APIs
MLOps & observability
Evaluation & reliability
Databases & infrastructure
Data, quantitative systems & tooling
08 / CONTACT

Contact

chinmayarvind23@gmail.com
Book a meeting LinkedIn GitHub Résumé

CHINMAY / NAVIGATOR

No destinations match “”.

Try a project name, “research”, or a technology like “PyTorch”.

0 destinations found.

navigate · Home / End open · Esc close