Ctrl + Shift + A to toggle
Personal ventures and notable engineering projects — from founding a product to deploying production infrastructure at scale.
Products and platforms I've built independently.
On-prem Media & Entertainment delivery and VFX ingest platform. Started as a fast parallel transcoder; grew into broadcast-grade QC, HDR/SDR color management, ProRes/DNxHR/DPX/EXR mezzanines, NLE round-trip across Resolve / Premiere / FCP / OTIO / EDL / AAF, a full VFX studio ingest pipeline with OCIO/ACES and Flow Production Tracking, RAW/CinemaDNG ingest, and AI-assisted preset selection — all running on hardware you already own. Built as solo technical founder.
Master/worker orchestration with five chunking strategies — closed-GOP slice, open-GOP remux, scene-aware, cut-aligned to AAF/EDL/FCPXML, and frame-range for EXR sequences. Cost-based scheduler with EMA-tracked runtime per worker × preset, GPU session budgeting, and work-stealing.
46 presets across 7 families with codec-spec conformance verification on every assembled output. HDR10 / HLG / SDR color metadata classification with zscale + tonemap chains. Deliverable-profile QC (LUFS + true-peak) against broadcast_atsc / broadcast_ebu_r128 / ott_netflix.
FCP-XML 1.10 + FCPXML + xmeml + OpenTimelineIO + EDL/AAF, plus opt-in direct Resolve project-DB linking. A hard backup-first rule snapshots every customer project file before any write — every relink is reversible.
EXR sequences via exrseq://, working-space + display transforms through oiiotool, ACES Studio Config v4 bundled. Shot domain model (Project → IngestProfile → Shot → ShotVersion). Flow Production Tracking with HMAC-SHA1 webhook + EventLog reconciler. Five-step ingest profile wizard.
Vite SPA served from the same origin as the API. Dashboard, Queue, tabbed job detail (Chunks / Workers / Logs / Spec / Forensics), Review (HLS preview), Ingest, Shots, virtualized Activity feed. Cmd-K command palette switches surfaces and themes. Simple and Advanced density modes.
JWT auth with approval-gated worker enrollment and HMAC-signed chunk URLs (60-min TTL). MAM push to Iconik / Dalet / Avid via a deliverable matrix. CLI for admin reset, DB backup/restore, firewall install, worker discover/probe/test-encode. Self-installing Windows firewall rules.
A persistent post-nuclear text MMO where every survivor on the map is a real person. Turn-based and server-authoritative, played in ten minutes a day while the wasteland grinds on without you — open PvP, a market run entirely by players, faction wars fought over days, and five skills you can never respec. Early access, free to play. Built as solo technical founder.
No NPCs to farm. Knock someone down and you choose what happens next — walk away, or pick them clean. Turn-based combat where every probability is shown before you commit: no hidden dice, no secret formula.
Going down isn't a respawn. You're left for dead with a triage timer counting and an injury that follows you — cracked ribs at −10% accuracy until someone patches you up. A stranger running Stabilize on you is a real player choosing to.
Every rebreather and round of ammo was scavenged, crafted, or looted by another player — and priced by them. Haul goods across the Rads between regions and eat the spread. List fees and sale tax burn out of circulation.
Every job shows success odds, payout range, stamina cost, body burden, and critical-fail chance before you attempt it. Blow the roll badly enough and a Tracker takes your scent.
Bind into a Faction, drill in its Cell, and put its name on the map. Multi-day wars tracked on war-points and momentum tiers — territory, rackets, and reckoning on a ledger everyone can see.
Muscle, Hardening, Reflex, Fieldcraft, Wits — one survivor, no wasted stats and no respec. Training costs Conditioning and accrues soreness. Ten regions to cross, and every road wants you dead.
Local-first AI desktop app for photographers, agencies, and stock teams — tags thousands of images with a vision-language model running on the user's own machine, then develops them in a Lightroom-style editor. Every byte round-trips with Adobe Bridge, Lightroom Classic, and NeoFinder. Built as solo technical founder.
Walks a folder, runs every supported image through a local vision-language model, and writes natural-language captions + 12–20 keywords into standard XMP / IPTC fields.
Exposure, tone curve, split toning, vignette, grain, crop, AI background removal, 2×/4× upscale, LaMa generative fill. Every slider writes to Adobe Camera Raw's crs:* XMP namespace.
Preview folders before importing — no DB writes, fast thumbnail grid, metadata-aware filter chips, rate / pick / reject keyboard shortcuts.
Opt-in InsightFace buffalo_l + sqlite-vec kNN — drop a reference photo, find every other photo of that person across workspaces. EULA-gated.
Browse, pull, and benchmark vision models from inside the app. Knows which models fit your VRAM and which licenses require explicit acceptance.
CSV, JSON, Lightroom XMP sidecars, NeoFinder XML, Markdown manifests, hierarchical Keywords.txt — plus AES-256-GCM encrypted .gvdb archives for Studio.
Browser-based professional audio platform — six studio-grade tools and a full audio engineering education system, built as solo technical founder from architecture through product.
A node-based visual processor where users build custom signal chains by connecting effects — compressors, limiters, noise gates, parametric EQ, filters, distortion, modulation, spatial effects, pitch/time tools, and mastering processors. Built on React Flow with a real-time audio engine.
State-of-the-art neural models separate any track into vocals, drums, bass, and instruments. Supports standard presets (vocals only, instrumental, 4-stem, 6-stem) and an advanced mode. Accepts MP3, WAV, FLAC, AAC, OGG, and Opus up to 100MB.
Intelligent mastering that matches a track to a professional reference — either a genre preset (Rock, Pop, Hip-Hop, Classical, Electronic) or a user-uploaded reference. Outputs WAV, MP3, or FLAC. Handles loudness targeting, dynamics, and spectral balance automatically.
Converts any audio recording into a playable MIDI file using AI pitch detection. Instrument-optimized presets for vocals, piano, guitar, bass, and mixed recordings. Supports batch processing of up to 5 files simultaneously and live microphone recording up to 15 minutes.
Preset-driven noise removal and voice processing covering podcast cleanup, heavy noise reduction, voice clarity, broadcast voice, and interview cleanup. Also includes music mastering and normalization categories. Processes any format up to 100MB.
21 modules, 263+ lessons, 205+ hours of content covering everything from sound physics to advanced DSP and genre-specific production. 13 core modules (fundamentals through EDM production) plus 8 genre specializations (Rock, Hip-Hop, Metal, Country, R&B, Post-Production, Jazz, Podcast). Includes 600+ quiz questions, XP progression, ear training, mix challenges, and 21 certificates. First lesson of every module is free.
Cloud rendering platform built on AWS Deadline Cloud — modern GPU infrastructure, DCC-native submitters, and transparent per-node-hour pricing for VFX professionals.
Significant engineering deployments and infrastructure builds at GPL Technologies.
Proof-of-concept engagement validating AWS Deadline Cloud for a full CBFX production pipeline — including cross-account RLM license connectivity and a hardened Rocky Linux 9 artist workstation image.
Production deployment of the full CBFX render farm on AWS Deadline Cloud — from infrastructure-as-code foundation through a Rocky Linux 9 artist workstation image used studio-wide.
Custom Service Managed Fleet worker build that brought licensed NeatVideo OFX plugins to Deadline Cloud — a solution AWS did not ship out of the box.
Production-ready CloudFormation template deploying a full Houdini 21 render farm on AWS Deadline Cloud — farm, queue, fleet, S3 attachments, IAM, and monitoring in a single stack.
POC proving out GPU rendering on AWS Deadline Cloud with two parallel queues — Cinema 4D 2026 with Redshift 2026.1, and Cinema 4D 2026 with C4DtoA (Arnold) — using Service Managed Fleets and Spot capacity.
Multi-phase AWS-funded engagement standing up a startup VFX studio on AWS — from network architecture through cost modeling, security, and delivery timeline — in partnership with AWS's Startup VFX program.
Proof-of-concept engagement moving JAMM VFX workloads onto AWS Deadline Cloud with a studio-specific SMF worker configuration and queue topology.
Proof-of-concept deployment validating Deadline Cloud as a render backend for Buck's motion design and VFX pipeline.
Proof-of-concept for Cantina Creative that had to beat on-prem render hardware on both speed and cost — benchmarked against a real 20K-resolution, 2,000-frame production sequence.
Proof-of-concept engagement deploying AWS Deadline Cloud infrastructure and validating end-to-end render workflows for Sight & Sound.
Production deployment of a custom Maya tile-render solution on AWS Deadline Cloud, enabling distributed tile-based rendering of oversized Maya frames across the SMF worker fleet.