Projects

Consulting engagements and independent software, from early AI infrastructure work to recent LLM and RAG systems.

#01

Sitespry

Built Sitespry — a RAG-powered document intelligence product for ingesting and querying unstructured content at scale.

  • Document ingestion pipeline for PDFs, Word files, and websites
  • Semantic embedding generation and auto-synthesised Q&A pairs
  • Hybrid vector + keyword search with inline citation references
  • Embeddable chat widget deployed via script tag
  • Plugin architecture for extending ingestion sources; MCP server integration

Python · FastAPI · Cohere · Gemini · libSQL/Turso · Cloudflare · Docker

#02

Zango.ai

Consultant developer for the core MVP of a regulatory intelligence platform — built to detect, analyse, and surface regulatory change for compliance teams.

  • LLM-based horizon scanning pipeline for continuous regulatory change detection
  • Document analysis pipeline: PDF parsing, embeddings, semantic search
  • AI chatbot for policy Q&A with grounded retrieval
  • Compliance gap analysis against internal policy documents

Python · FastAPI · PostgreSQL/pgvector · Claude · Vertex AI · GCP · Firebase

#03

Earnr UK

Designed and built an open-book QA assistant for answering UK tax and accountancy questions, for a London-based fintech startup.

  • Hybrid retrieval system combining keyword and semantic search across internal documents, customer data, and public tax documentation
  • Custom web scrapers for embedding data from external sources
  • GPT-4 function calling router for questions requiring tax computation, dynamic diagrams, and illustrations; schema validation with Pydantic
  • Ensemble re-ranker adapted from llama-index and LangChain
  • LLM routing across OpenAI, Perplexity, and open-source models for rate limiting, cost, and latency optimisation
  • Released as a ChatGPT plugin
  • DuckDB analytics system for observability

Python · Pydantic · pgvector · lanceDB · DuckDB · RAG · OpenAI · Perplexity

#04

Taloflow (YCombinator 2021)

Backend infrastructure and data tooling for a YC-backed cloud intelligence platform.

  • Migration of a large Postgres cluster from self-hosted to AWS Aurora, with query optimisation and production scaling
  • Role-based access control using Postgres RLS, JWT, and PostGraphile middleware for GraphQL
  • Backend API and internal UI for managing a data catalog: AWS AppSync, Lambda, MongoDB, PostgreSQL, SQLAlchemy, FastAPI
  • Benchmarked object storage providers: S3, GCS, Wasabi, Storj DCS

Python · FastAPI · Pydantic · PostgreSQL · Google Cloud Run · AWS App Runner · DynamoDB

#05

Autotrip

Founding engineer and CTO/tech lead at an automatic mileage tracking startup — built the core platform from scratch.

  • Automatic mileage tracking and business expense classification using custom hardware sensors and mobile apps
  • IoT platform for handling streaming sensor data from company car fleets
  • End-to-end system: data ingestion, storage, processing, and driver-facing apps

Python · SQLAlchemy · PostGIS · AWS Lambda · S3 · Flask · PostgreSQL · Docker · Twilio · IoT · GPS

#06

Sustainable Ventures

Full-stack developer embedded with early-stage startups in the green energy and sustainability sector, helping founders validate ideas and ship MVPs with minimal cost.

  • Competitive research, rapid prototyping, and MVP builds across multiple companies
  • Data pipelines, monitoring dashboards, and infrastructure for greentech products

Python · InfluxDB · PostgreSQL · AWS Lambda · EC2 · Docker · Grafana · Redis

#07

geolib

Python port of Chris Veness' geohashing algorithm — a geospatial utility library published on PyPI and used in geolocation and proximity applications.

  • Geohash encoding and decoding, bounding box queries, and neighbour lookups
  • Zero dependencies; designed for use in backend APIs and data pipelines

Python · PyPI · Open source — pypi.org/project/geolib

#08

geofilter.dev

An AWS Marketplace service for transforming streaming geospatial data from IoT devices, GPS sensors, and mobile apps into geo time series, trip segments, route maps, and address elements.

  • Real-time geospatial stream processing as a managed service
  • Built and sold independently as a bootstrapped SaaS product
  • Deployed on AWS with full documentation and API

Python · PostgreSQL · PostGIS · FastAPI · Docker · AWS · CloudFlare Workers · Open source

#09

is-aws-down

Continuous health check monitoring for AWS services — an open-source tool built with Taloflow to provide real-time visibility into AWS outages and degradations.

  • Polls AWS service health endpoints and surfaces status in a public dashboard
  • Multi-region checks across core AWS services
  • Deployed on Google Cloud Run and AWS App Runner for independent redundancy

Python · FastAPI · Google Cloud Run · AWS App Runner · DynamoDB · PostgreSQL — github.com/Taloflow/is-aws-down

Start a conversation →