Selected Work

Work that connects AI systems to real business operations.

Start with the HeySynth case study: a production AI story spanning forecasting, cloud workers, backend APIs, agent orchestration, and operator-facing React surfaces. The lab builds below are smaller experiments that show how I prototype and explore new AI interaction patterns.

Problem

What was hard, who felt the pain, and why AI was worth introducing.

System

The model, agent, API, data, and frontend decisions behind the product.

Outcome

What shipped, what improved, and what the work changed for users or the business.

Evidence-backed case study

Case study 01

HeySynth

Building the intelligence layer for retail teams managing real inventory decisions.

A story-led case study on how I helped turn forecasting, agent workflows, cloud worker reliability, backend APIs, and React surfaces into production-oriented retail planning systems at Techstars-backed HeySynth.

Role

Fullstack AI/ML Engineer

Scope

Forecasting engineAgent orchestrationCloud workersMLOps reliabilityBackend APIsOperator UI

The hard part was not just producing forecast numbers. Retail operators needed forecasts they could control, audit, override, and trust across different channels and SKU histories.

My work connected the intelligence layer end to end: forecasting logic, Forecast Mode APIs, cloud-style worker pipelines, LangGraph-style agents, and the React surfaces operators used every day.

The result was a more production-minded planning system: safer model fallbacks, controllable risk posture, cleaner AI outputs, and MLOps reliability around long-running forecast workflows.

Open case study

Codebase surfaces: ML, worker, backend, agents, frontend

Workflow families across forecasting, agents, imports, analytics, chat, and collaboration

Focused Deep Dives

Two HeySynth stories separated into sharper case studies.

HeySynth is one product ecosystem, but the work has multiple distinct systems inside it. These focused pages break the work into two clear lenses: production forecasting and agentic operator workflows.

Forecast Mode: making demand forecasting controllable for retail teams. preview

Evidence-backed deep dive

Focused deep dive

HeySynth

Forecast Mode: making demand forecasting controllable for retail teams.

A focused technical case study on forecasting strategy, model routing, channel/SKU controls, import reliability, and the MLOps path around long-running forecast workflows.

ML forecastingForecast Mode controlsMLOps reliabilityImport pipelinesPlanning UI

Focused case study available

Open the focused deep dive with real product evidence, implementation narrative, and technical decision context.

Open focused case study

Ask Synth + Flow: turning retail AI agents into daily operator workflows. preview

Evidence-backed deep dive

Focused deep dive

HeySynth

Ask Synth + Flow: turning retail AI agents into daily operator workflows.

A focused product-agent case study on Ask Synth, Flow, Playbook, LangGraph-style orchestration, tool-calling guardrails, output cleanup, and operator-facing decision support.

Agent orchestrationAsk SynthFlowPlaybookOutput safety

Focused case study available

Open the focused deep dive with real product evidence, implementation narrative, and technical decision context.

Open focused case study

AI Labs

Prototype builds and technical experiments.

These are smaller builds, demos, and explorations across realtime AI, agent orchestration, retrieval, mobile workflows, and data products. They are not positioned as production case studies; they show technical range, speed, and curiosity.

Demo build

Vocalis AI

Ultra-low latency real-time voice assistant handling continuous conversations with sub-500ms response times. Features live waveform visualization, interruptibility, and emotion detection using OpenAI Realtime API and WebSockets.

Next.jsOpenAI Realtime APIWebSocketsReact Audio VizEdge Functions

Open demo

Prototype

AgentGraph Studio

A fully functional visual orchestrator for multi-agent LLM workflows. Features interactive "Agent Nodes" with customizable system prompts (e.g., "Sarcastic Poet"), real-time model switching (GPT-4 vs GPT-3.5), and live execution visualization with feedback loops.

React FlowOpenAI APIZustandShadcn UINext.js

Open demo

Build

Sentinel

On-demand security dispatch platform (Uber-for-Security). Clients can request protective agents for specific missions. Features real-time geospatial tracking, agent matching, and WebSocket-based mission status updates.

FlutterGoogle Maps APIFirebaseDartGeofencing

APK iOS access

Demo build

Christosin.com

Digital media distribution platform for books and audio. Integrating a RAG layer to enable "Semantic Query" across the library, allowing users to converse with book contents for deep theological insights.

Next.jsPostgreSQLVector DBOpenAIStripe

Open demo

Experiment

InspireEdge AI

Lead Engineer for an AI-powered commerce intelligence engine. Architected the entire ML pipeline to detect cart abandonment and predict user LTV in real-time. Reduced abandonment by 22% for pilot clients through automated, hyper-personalized interventions.

PythonFastAPIRedisLangChainSupabase

Open demo

Experiment

Aircraft Engine Valuation Pipeline

Built a fault-tolerant extraction pipeline converting unstructured maintenance logs (PDF/Scans) into structured valuation data. decreased manual data entry time by 94% using a hybrid approach of regex heuristics and LLM validation.

PythonGroqGPT-4OCRData Engineering

Open demo

Experiment

Finsight AI Terminal

Real-time financial analytics dashboard aggregating news, sentiment, and price data. Features a custom anomaly detection engine and "Chat with Data" interface for traders, handling 50+ simultaneous data streams.

Next.jsFastAPITimescaleDBWebSocketsRecharts

Open demo

Experiment

Crypto Sentiment Agent

Production-grade NLP pipeline classifying cryptocurrency news for "Hype" vs. "Utility" signals. Deployed custom fine-tuned Transformer models on Hugging Face Inference Endpoints for sub-100ms latency.

PyTorchHugging FaceTransformersReactMLOps

Open demo Code

Work that connects AI systems to real business operations.

Problem

System

Outcome

Building the intelligence layer for retail teams managing real inventory decisions.

Two HeySynth stories separated into sharper case studies.

Forecast Mode: making demand forecasting controllable for retail teams.

Ask Synth + Flow: turning retail AI agents into daily operator workflows.

Prototype builds and technical experiments.

Vocalis AI

AgentGraph Studio

Sentinel

Christosin.com

InspireEdge AI

Aircraft Engine Valuation Pipeline

Finsight AI Terminal

Crypto Sentiment Agent

Let's build useful AI systems, not impressive demos.