35%
Fine-tuning cost reduction with CloudTune
About
Systems-minded founder energy
I am a New York University CS + Math student obsessed with the infrastructure layer: I build multi-cloud GPU brokers, agentic AI backends, compliance-grade audit systems, wireless measurement pipelines, and trading engines that can survive production constraints. My week is split across NYU Wireless research, Vertically Integrated Projects, and founder sprints for CloudTune, Axiom, and a few venture experiments.
Tooling sweet spot: Python · FastAPI · Go · Java · Terraform · AWS/GCP/Azure · Docker/Kubernetes · Redis/PostgreSQL/MySQL · PyTorch/CUDA plus COLMAP/Open3D, NYUSIM/NYURay, and Grafana/Prometheus observability. I like to prove ideas with research rigor, then wrap them in an ops-friendly platform.
- Multi-cloud orchestration + GPU scheduling (CloudTune)
- Agentic AI backends with receipts & policy rails (Axiom)
- 3D environment + FR3 datasets for wireless research (NYU Wireless)
- Compliance-ready logging + quant infrastructure (algo trading + RevSplit)
- Productized student ventures (ChessBet, Certify, LabScoutr)
NYU CAS · CS major · Math minor
NYU Wireless · Prof. Ted Rappaport & Prof. David Fouhey
VIP AI in Education · OptiReality
Greene & SIM supercomputing
CloudTune founder
Axiom co-builder
Think Do Thrive grant recipient
Flagship Builds
Products that feel like startups
CloudTune — No-code multi-cloud LLM platform
One-click GPU broker for LoRA/QLoRA fine-tuning. Terraform-driven orchestration across AWS, GCP, Azure Container Apps, and Lambda Labs with autoscaling FastAPI services. Built dataset ingestion UI, TensorBoard streaming, artifact receipts, and full-stack observability (Prometheus/Grafana + traces) that cuts MTTR dramatically.
- 20-35% cheaper fine-tuning via quant-aware training, KV cache pruning, and spot scheduling.
- GPU-aware scheduler, receipt ledger, and cost attribution for every training run.
- Front-end: React + Tailwind dashboard for uploads, receipts, and live metrics.
Axiom — AI compliance & audit microservices
Compliance OS for AI and trading teams: WORM log writer, notarized receipts, binder generation, and policy-as-code guardrails. Architected as a mesh of Go / FastAPI services with Redis streams and PostgreSQL event sourcing.
- Real-time policy evaluation + SOC2-lite evidence graph across cloud regions.
- Auto-generated audit binders, lineage explorer, and integration with Cloudflare Workers.
- Used by early fintech teams to satisfy SEC/FINRA retention expectations.
Multi-layer crypto & equities algorithmic trading engine
Low-latency ingestion from Coinbase WebSockets, real-time order book pressure modeling, and RL alpha strategies (SAC + PPO) running on NYU Greene. Deployed Grafana dashboards for risk, plus backtest harness with Hydra configs and vectorized simulations.
- Live state machine coordinating execution, hedging, and kill switches with 99.9% uptime.
- Feature store backed by Redis + DuckDB for microstructure signals and synthetic data.
- Latency-optimized Python/C++ stack with CUDA kernels for heavy analytics.
Ventures
Experiments with founder-market fit
Peer-to-peer wagering for online chess
A YC-style MVP that lets verified Chess.com / Lichess players spin up real-time wagers backed by an escrow wallet and instant dispute tooling.
- Anti-cheat layer that inspects PGN streams + engine drift.
- FastAPI + WebSockets backend with event-driven payouts.
- ELO-aware matchmaking plus compliance checks per match.
Searchable certification intelligence
Ranked catalog of CS, finance, and cybersecurity certifications so students can compare difficulty, cost, and study paths like a college-search portal.
- Elastic-style search with filters for modality, provider, and stack.
- Difficulty tiers, salary lift data, and curated prep tracks.
- SEO-first content engine for organic growth.
$2K Think Do Thrive winner for research matching
Rate-my-lab marketplace built with the Burton D. Morgan Foundation + Veale Institute; surfaces mentorship style, culture, and workload across labs.
- Students review PIs, equipment access, and publication cadence.
- Search + filters across disciplines with trust + anonymity levers.
- Funds go to onboarding NYU + Case Western pilot cohorts.
Split-tracking intelligence for retail investors
Python + Alpaca/Polygon powered feed that tracks upcoming stock splits, reverse splits, and the downstream liquidity impacts.
- Event calendar with alerts + brokerage-grade CSV exports.
- Historical models explaining price behavior post-event.
- Educational explainers to onboard newer investors.
Research & Academic Work
Applying AI to wireless systems, perception, and learning
Comprehensive FR3 Radio Propagation Modeling
NYU Wireless · Prof. Theodore S. Rappaport · Nov 2023 - May 2024
- Executed upper-mid band campaigns (6.75 / 16.95 GHz) across 100+ TX-RX pairs with standardized capture + documentation.
- Unified ~150 GB of power-delay profiles + metadata into a NYUSIM-ready dataset for global researchers.
- Leading publication efforts that benchmark FR3 reliability for future 6G systems.
Material-aware 3D Scene Reconstruction for Ray Tracing
Prof. David Fouhey & Prof. Rappaport · Sep 2023 - Present
- Reconstructed ~2,500 sqm indoor space with <=10 cm geometric error and 89% material precision, improving NYURay by +18%.
- Extended to 2 sq km outdoor captures using Record3D + COLMAP with plane/object segmentation.
- Pipeline powers wireless ray-tracing simulations and robotics experiments.
VIP: AI in Education & OptiReality
NYU Vertically Integrated Projects · 2023 - Present
- Deploy AI copilots that translate lecture signals into adaptive feedback for instructors.
- Prototype OptiReality overlays that merge LiDAR maps with vision-language grounding.
- Coordinate cross-discipline student teams; present findings to NYU faculty sponsors.
Experience
Teaching, backend platforms, and infrastructure
Teaching Assistant — Introduction to ECE (ECE-1002)
Feb 2025 - May 2025 · NYU WirelessMentored 160 students through labs + office hours, authored ABET-aligned assignments, and co-wrote a new ECE/CS fundamentals text with Prof. Rappaport.
MentorshipEmbedded labsCurriculum design
Research Engineer — NYU Wireless
Nov 2023 - Present · NYU Wireless / NYURayLed FR3 (6.75 / 16.95 GHz) campaigns, curated 150 GB of PDP data, and reconstructed 2,500 m2 indoor + 2 km2 outdoor environments with COLMAP, Record3D, and Open3D to improve NYURay accuracy by 18%.
FR3COLMAPOpen3DNYUSIM
AI-Driven Backend Systems Intern — Nanjing Tech Company
Jan 2023 - Dec 2023 · Nanjing, ChinaLaunched an AI Ops control plane (Java, Spring Boot) that cut down time 15%, boosted throughput 40% via multithreading + Redis caching, and trimmed latency 25% through MySQL tuning—all while holding 99.9% uptime.
JavaSpring BootRedisMySQLObservability
Founder / Platform Engineer — CloudTune & Axiom
2024 - Present · NYCBuilding venture-scale infra as a student: compliance rails for regulated AI teams plus a multi-cloud GPU broker that ships modelformation as easily as shipping code.
FastAPITerraformAWS / GCP / AzureReactPrometheus
Focus Areas
How I operate
AI Systems
LoRA/QLoRA, RL, vector DBs, retrieval, model evaluation, inference cost modeling.
Distributed Infra
Autoscaling microservices, Terraform modules, multi-cloud GPU scheduling, SOC2-lite controls.
Wireless & Sensing
FR3 measurements, LiDAR/COLMAP recon, NYUSIM modeling, ray-tracing friendly datasets.
Algorithmic Markets
Market microstructure signals, RL trading agents, kill switches, observability pipelines.
Stacks & Skills
Full-stack founder toolchain
Languages
Python, Java, TypeScript/JavaScript, C/C++, SQL, CUDA
AI / ML
PyTorch, TensorFlow, RL (PPO, SAC, DQN), LoRA + QLoRA, vector databases, Hugging Face pipelines
Cloud & DevOps
AWS (SAA-Pro + Security), GCP PDE, Azure, Terraform, Docker, Kubernetes, Redis, PostgreSQL, MySQL, Grafana, Prometheus, Cloudflare
Systems & Specialties
Distributed systems, GPU scheduling, API design, observability, wireless communication, 3D reconstruction, algorithmic trading
Coursework
Curriculum aligned with the work
CS & Software
- Basic Algorithms
- Operating Systems
- Computer Systems Organization
- Software Engineering
- Data Structures & OOP (Java / C++)
- Computational Tools for Engineers
ML, AI & Data
- Fundamentals of Machine Learning
- Natural Language Processing
- Statistical Learning Methods
- Deep Learning Systems (PyTorch/TensorFlow)
- Reinforcement Learning for Wireless Systems
- Data Science for Engineers
Cloud & Infrastructure
- Distributed Systems
- Database Systems (SQL/PostgreSQL/MySQL)
- Cloud Computing on AWS/GCP
- DevOps & Infrastructure as Code
- Parallel & High-Performance Computing
Wireless & 3D Vision
- ECE Fundamentals
- Signals & Systems
- Wireless Communications (FR1-FR3)
- Electromagnetics & Propagation
- 3D Vision & Reconstruction (COLMAP / Open3D)
- Ray Tracing for Wireless Systems
Mathematics
- Calculus I-III
- Linear Algebra
- Probability & Statistics
- Discrete Mathematics
- Optimization Methods
Writing
Deep dives, build logs, and research notes
CloudTune: Distributed GPU Broker Playbook
Lessons from turning CloudTune into a one-click GPU broker: Terraform modules for AWS/GCP/Azure, FastAPI schedulers that respect spot volatility, and the observability stack that keeps LoRA / QLoRA training receipts compliant.
Read essayAxiom: Compliance Rails for Fast AI Teams
How we designed WORM evidence stores, microservice guardrails, and binder automation so regulated AI + trading teams can ship product without waiting on legacy governance vendors.
Read essayRealtime Trading Engine Architecture Notes
A breakdown of my multi-layer Coinbase engine: order-book pressure features, RL agents (SAC/PPO), Grafana dashboards, and the kill switches that keep student trading infrastructure sane.
Read essayMore essays + build notes live on LinkedIn.
See LinkedIn HighlightsCerts & Highlights
Proof I obsess over infra
Certifications
- AWS Solutions Architect — Professional & Associate
- AWS Security — Specialty
- Google Professional Data Engineer
- Oracle Java SE 11 Developer
Programs & Activities
- NYU VIP: OptiReality · AI in Education · FAMS
- NYU Wireless Research Group
- Google + NYU hackathons & accelerator programs
- Leslie eLab / Startup Sprint initiatives
Grants & Recognition
- Think Do Thrive Grant - $2,000 for LabScoutr
- NYU Wireless Research recognition for FR3 + NYURay work
- NYU Greene / SIM allocations for FR3 + trading workloads
Contact
Let's build something audacious
I'm always down to jam on ML infrastructure, GPU scheduling, wireless research, or trading systems. Reach out for internships, collabs, or startup sprints.