/ case studies
Selected work.
A small, honest sample of the work in the last 24 months. Detail and code samples available under NDA.
YouTube Shorts automation, local-only.
The client needed a local-only YouTube Shorts automation system — topic planning, script generation, scene-split, image + voice generation, FFmpeg-based video rendering, and YouTube upload with scheduling — all running on a single developer machine using Next.js as the sole backend (no separate API server). PostgreSQL with Prisma persisted topic batches, scenes, render status, and […]
AI-augmented MT5 scalping bot.
Manually trading scalps during off-hours meant missing the windows that mattered most. The goal was to keep the operator in the loop — every trade decision logged with its AI reasoning and indicator state — while letting the bot act on windows the operator wasn’t watching. Our team built a Windows-friendly bot that connects to […]
Hybrid retrieval RAG pipeline.
Traditional keyword search finds files that *might* contain the answer; the user then opens them and reads through. Our pipeline finds the specific passage, reads it, and gives back a direct answer with a citation the user can open to verify. Real uses from the project’s own README: ingest a contract and ask “who owns […]
Quantitative trading research platform.
Quantitative research lives or dies on reproducibility: same data slice, same strategy, same backtest result, every time. Our team scoped and shipped the platform foundation: a Next.js 15 dark-themed front-end (TanStack Query v5 for server state, Zustand for client state, Recharts for financial charts, React Hook Form + Zod for forms) over a FastAPI async-first […]
Cross-exchange orderbook aggregator.
Trading on one exchange and watching another for arbitrage means juggling four browser tabs, four API keys, and four rate-limit budgets — and missing the prints that matter in the gaps. Our team built the aggregation service as a single Bun binary with no external runtime dependencies. The 13KB browser bundle runs the same aggregation […]
Prediction-market resolver.
Existing prediction-market oracles have three structural flaws: resolution becomes truth by inertia if no one challenges a stale proposal within the short window; challenging costs a bond that prices out small holders; and final disputes are decided by token weight, not evidence. Our team shipped an alternative: an AI agent bound by the market’s own […]
Telegram Mini App on TON.
Telegram Mini Apps live or die on three things: the WebApp SDK has to be present at runtime (we mock it in dev so the same code runs without Telegram injected), the game loop has to be rewarding within a single tap (mining rate visible immediately, first reward within seconds), and the leaderboard has to […]
Smart-contract risk monitor, 16 chains.
The original linux install was lost when that machine was wiped, so we reconstructed everything from the surviving session logs of two production cron runs. The key insight those logs proved: cursor persistence was decoupled from analysis. The cursor advanced and was written to disk *before* the slow Slither/ML stage ran. A crash or timeout […]
Solana trading platform, end-to-end.
Solana memecoin windows move in minutes — by the time a single monolith finishes ingesting a Telegram signal and running a rug-check, the trade window is gone. Splitting into four focused services (alpha-radar signals, copy-pulse mirroring, trade execution, shared types) let signal ingestion run at message rate while execution stayed independent. Every signal and every […]
Real-estate tokenisation platform.
The client needed institutional-grade tokenisation of property portfolios — every share issuance, secondary transfer, and corporate buyback had to be traceable from on-chain event to investor statement. Off-chain bookkeeping had to keep up with on-chain velocity, including a same-day reset after a portfolio-restructure incident mid-launch. Our team scoped, designed, and shipped the entire stack: EVM […]