Aaryan Dhandsoftware Engineer

An on-demand tow marketplace for Alberta. A stranded driver is matched to a nearby verified operator in under two minutes, at a locked price. I own the product direction and build the driver app and dispatcher dashboard.

• PRD

  Author the product requirements and hold the scope line: four client surfaces (customer app, driver app, web booking flow, dispatcher dashboard) sequenced for a three-founder team against a hard pre-winter launch, with the cut-order agreed in advance for when the sprint slips.

• D1

  Made analytics a day-one requirement rather than a later migration: a lightweight warehouse synced off the operational Postgres, so pilot success criteria (acceptance rate, driver retention, unit economics) are measured as they happen instead of reconstructed after the fact.

• 4.5

  Replaced a raw star average with a Bayesian rating that regresses toward a prior until a driver has enough ratings to be meaningful. At pilot volume, one early review would otherwise misrepresent an operator's quality and trigger a review gate that isn't statistically real yet.

• GTM

  Hold the team to a supply-first launch: no customers until verified operators with confirmed idle capacity are live and have tested the app, gated on a written go/no-go. Run driver discovery, chair the weekly sync, and own the Jira boards, Git workflow and release process across three founders and an operations lead.

• React Native
• Expo
• Next.js
• Node.js
• PostgreSQL + PostGIS
• Supabase Realtime
• Stripe Connect

Real-time interpretation of biometric signal, shipped inside a live consumer health app, and the serving infrastructure underneath it.

• 57%

  Cut LLM inference latency by re-architecting the Anthropic-powered insight layer into a retrieval-augmented (RAG) pipeline that pulls the prior week of time-series biometric context into the model's window to ground real-time interpretations.

• 0–1

  Built and deployed a semi-supervised multimodal stress model fusing HRV, Apple Health biometrics, chat, and mood check-ins into a calibrated stress score, served in real time in a live Flutter + Firebase iOS app.

• →

  Moved research prototypes into a production analytics product by deploying TensorFlow/PyTorch models behind RESTful APIs and building the data-preprocessing and model-serving infrastructure they run on.

• Anthropic API
• RAG
• PyTorch
• TensorFlow
• Flutter
• Firebase
• REST APIs

A centralized warehouse, the pipelines that fill it, and the dashboards that finally made a utility's cost data legible in real time.

• SQL

  Built the data foundation for downstream ML and predictive analytics, standing up a centralized Microsoft SQL Server warehouse ingesting SAP and OPC data via optimized SQL/Python pipelines with data-quality validation.

• 6×

  Accelerated material quantification by leading a 6-person team to design scalable Python/VBA ETL workflows over 300+ historical project records, laying groundwork for forecasting and machine learning.

• 95%

  Cut manual reporting workload by engineering end-to-end automation across SharePoint, SAP, Excel VBA, and Python for baseline and month-end cost reporting, documented for handoff.

• BI

  Deployed interactive Power BI dashboards wired to live SAP data via custom Python and Power Query pipelines, giving engineering and project leaders real-time cost and inventory visibility.

• SQL Server
• Python
• SAP / OPC
• ETL
• Power BI
• Power Query
• Excel VBA