Software Engineer · Backend Systems · Applied AI
Building secure, scalable infrastructure for decision intelligence.
Featured Project
Smart Strategies Builder
AI-powered financial decision intelligence platform
A production-ready financial decision intelligence platform focused on risk analysis, simulation, and explainable insights — not trade execution.
Impact
- Designed and built a full-stack SaaS platform using FastAPI, PostgreSQL, Redis, Docker, and Next.js, with secure authentication, subscription billing, and role-based entitlements.
- Implemented a tiered intelligence access system enforcing feature availability by subscription (risk analytics depth, data latency, simulation limits) with Redis-cached entitlements.
- Architected a simulation-first, explainable approach to financial intelligence, emphasizing risk metrics, regime analysis, and backtesting — explicitly avoiding live trading or profit claims.
Tech Stack
What This Demonstrates
- Production API design (auth, billing, webhooks, entitlements)
- Data-driven feature gating and access control
- Clear separation between intelligence and execution
- Readiness for ML/AI systems (evaluation, explainability, safety)
Key Features
- Secure auth (JWT, MFA, Argon2id)
- Stripe subscriptions + webhooks
- Intelligence entitlements by tier
- Portfolio risk analytics and simulations
- Backtesting and historical replay
- Audit-friendly, explainable outputs
SSB is a financial decision intelligence platform. It does not execute trades, generate buy/sell recommendations, or promise returns.
Other Projects
ARIA
Self-hosted AI agent runtime
A personal AI agent that runs entirely on my own hardware. FastAPI server with an agentic tool-use loop, local inference on an RTX 5060 Ti via Ollama, and a provider abstraction that swaps between local and hosted models through config alone. Tools run inside a bubblewrap sandbox with no network and no access outside the workspace. Context compaction keeps long conversations inside a 16k window without losing the transcript. 198 tests, strict mypy, no model-specific workarounds in the agent layer.
Tech Stack
Highlights
- Provider-neutral agent loop: Ollama and Anthropic tool-call formats are translated inside each provider, with zero model-specific branching above that layer
run_pythonexecutes inside a bubblewrap mount namespace with--unshare-net, no$HOME, and only the workspace bound, proven by an adversarial escape suite, not asserted- Sandbox probe runs at startup and refuses to execute rather than silently falling back to unconfined
- Context compaction is a projection, not a deletion: the canonical SQLite transcript is never mutated, only the per-request view sent to the model
- Per-session turn guard closes a load-run-append race that would otherwise interleave two concurrent requests into one incoherent transcript
- 198 tests, strict mypy, zero warnings
FinRAG — Finance Domain RAG System
Production-grade Retrieval-Augmented Generation system for financial document intelligence.
A finance-domain RAG system built to analyze SEC filings, earnings reports, and financial documents through grounded LLM responses. Demonstrates full-stack ML engineering ownership across ingestion, chunking, embeddings, vector search, reranking, and API deployment.
Tech Stack
Highlights
- Secure document ingestion pipeline with validation + sanitization
- Finance-aware chunking strategy
- OpenAI text-embedding-3-small for retrieval, GPT-4o-mini for grounded answer generation
- Custom MMR reranker for result diversity
- Vector DB abstraction layer
- Async LLM querying
- Strict API validation + rate limiting
- Structured logging with Structlog
- Production security architecture
Discern
Computer vision that turns paper records into structured, searchable data.
A deployed, full-stack document-intelligence app that digitizes paper records into structured, searchable data. Upload a photo or scan — Discern detects fields, runs OCR, and returns editable results with confidence scores, built end-to-end with PyTorch, FastAPI, and Next.js.
Tech Stack
Highlights
- Synthetic training data generation for handwritten form fields
- Detection + OCR pipeline for church connection cards and prayer requests
- Evaluation harness for measuring model accuracy
- FastAPI service exposing structured data endpoints
- Next.js UI for search and browsing digitized records
- PostgreSQL schema for structured, queryable output
- Fully containerized with Docker
Client Work
Freelance builds for real businesses and organizations — scoped with the client and built for production.
Presidents Closet
Streetwear e-commerce brand
Moved a live e-commerce store off the Wix page builder onto a custom Next.js storefront built from scratch. Product and inventory data still come live from Wix, and cart and checkout process real customer payments.
- Integrated the Wix Headless API so the client keeps managing products in Wix while customers shop on the custom frontend
- Owned the full domain migration: DNS cutover from Wix hosting to Vercel for the production domain
- Shipped a working cart and checkout flow handling real orders
Tech Stack
Positive People
Daily positivity & mindset app
A full product build for a founder: an installable progressive web app (PWA) with daily content and community features, plus an admin dashboard so the client can run it without a developer.
- Shipped as a PWA so users can install it straight from the browser, with no App Store or Play Store review
- User accounts, onboarding, a daily content system, mood check-ins, streaks, and points/rewards to keep users coming back
- Private journal, community feed, and a video section, all managed from an admin dashboard
Tech Stack
Nia Imani Academy
Non-profit education organization
A full website rebuild for a non-profit, plus Stripe integration so the organization can take donations, tutoring fees, and camp registrations online.
- Rebuilding the public site while the current one stays live
- One Stripe integration covering three revenue streams: donations, tutoring, and camp registration
Tech Stack
Greater New Bethel Church
Church & community organization
A new website for a local church, with a clean, warm, community-focused design that reflects the congregation.
Tech Stack
About Me
I'm a software engineer focused on building production systems that are secure, scalable, and built to last.
My work spans full-stack development, backend infrastructure, and applied AI — with a bias toward systems that require real architectural thinking: auth, billing, data pipelines, and decision intelligence platforms. I care about code that ships, systems that don't break at 3am, and abstractions that survive contact with production.
I studied Computer Science, where I built the foundation — Java, Python, C++, algorithms. Since then, I've been sharpening on real projects: designing APIs, implementing secure auth flows, and building SaaS infrastructure from the ground up.
I'm looking for a role where I can own meaningful backend systems, contribute to ML/AI-adjacent products, and work alongside engineers who take craft seriously.
Skills & Technologies
Languages
Backend
Frontend
DevOps
Security
ML / AI
Contact Me
Email: standleydowell@gmail.com