Founder / freelance full-stack product engineer
Freelance software development, technical consulting, and founder-led product work across Superscribe, x11.social, AI voice workflows, trained speech models, production automation, business intelligence, data engineering, domain-specific ML, APIs, backend systems, frontend clients, self-hosted infrastructure, and workflow tooling.
ERP automation and domain models
Turned quoting, orders, worksheets, production planning, worklogs, machine events, warehouse/materials, compliance workflows, integrations, and AI-assisted email-to-bidding into one operations system. Trained domain-specific models for paper and format matching, then blended those predictions into agent-driven automation flows.
Business intelligence and reporting
Built reporting layers, analytics views, business metrics, and operational visibility for production workflows, orders, worklogs, transactions, and internal decision-making.
AI training-data pipelines
Prepared, cleaned, joined, and structured datasets for AI model training, including feature extraction, domain-specific training sets, and evaluation loops for paper/format matching, voice, scoring, and trading-model experiments.
AI sales voice and model work
Built private sales-call automation from outbound dialing to live transcription, speech detection, LLM turn-taking, streaming TTS, recordings, Estonian ASR experiments, and voice synthesis/model training.
Trading research and backtesting lab
Built a trading-bot research stack around tick bars, order-flow features, Wyckoff-style setup detection, walk-forward/backtest validation, and Mamba-style sequence models. Validation artifacts include long-side AUC above 0.73 and broader risk/expectancy experiments.
X scorer model and Chrome extension
Built a local-model Chrome extension for scoring X/Twitter posts and composer drafts, combining virality, clarity, novelty, shareability, authenticity, slop/clickbait/rage/context signals, feed badges, composer hints, and local ONNX inference.