Anna Poutanen

Software Engineer — Mercor

Sept. 2026 – Present

• Build Vibe Code Bench, a 2,000+-task benchmark for frontier coding agents featured in APEX Benchmarks.
• Built layered verifiers using API contract tests, PostgreSQL state checks, and rubric-scored headless-browser flows.
• Built a pipeline running 10+ coding agents in parallel to repair and deliver 130+ benchmark tasks, with automated quality checks and CI failure recovery.

Software Engineer & Project Manager — Ascendance Foundry

Aug. 2026 – Present · Toronto, ON

• Independently built backend APIs and PostgreSQL schemas on Google Cloud for a trading platform supporting US$400K in invested capital and 1M financial records.
• Led the trading-platform project with 3 engineers, coordinating frontend and stock-evaluation development and integrating their components with the backend APIs.
• Built a stock-valuation engine for 250 equities, cutting held-out valuation error by 18% versus a Claude baseline.

Forward Deployed Engineer — Ascendance Foundry

May 2026 – Aug. 2026 · Toronto, ON

• Designed a recall-first PII-responsiveness classifier in Python for breach-notification review that triaged 208K documents, achieving 97% recall and a 1.95% missed-clearance rate, cutting first-pass manual review by about 47%.
• Standardized the tool for reuse across breach-review projects, achieving over 95% recall on sample sets from 6 previous projects; added a web interface for running reviews and viewing results.
• Shipped an end-to-end AI prospecting tool in Python (firm analysis, contact sourcing, personalized outreach, automated sending, CRM logging) that matched the vendor tool it replaced at 1/80 of the cost.

Data Analyst — Mount Sinai Hospital

Feb. 2025 – Aug. 2025 · Toronto, ON

• Created a pilot quality-control dataset encoding 429 validation rules to confirm rule behaviour before running against production lab tables of hundreds of thousands of entries; the pilot caught 94.8% of the errors found in the full dataset.
• Automated sorting and analysis of lab data by type against the same rule set in Python (pandas, NumPy, Git), replacing manual spot-checks.

Math Tutor — Kumon Tutoring Center

May 2025 – Aug. 2025 · Toronto, ON

• Tracked student progress, delivered individual feedback, and supported daily class operations.

Coding Camp Counselor — UCC Summer Camps

June 2023 – Aug. 2023 · Toronto, ON

• Taught Python fundamentals to students ages 12–15 through week-long lesson plans and interactive projects.

Coaching & Officiating

2022 – 2025 · Ontario

• Assistant Volleyball Coach, Titans Volleyball Association (2025).
• Soccer Referee, Beach Community Soccer League (2022–2025).
• Volleyball Camp Counselor, Camp Madawaska (2024).

PgDog — Open Source

Oct. 2026 · Rust, PostgreSQL

• Fixed SET LOCAL … DEFAULT handling in Rust, preserving session timeouts and warnings outside transactions; tested simple/extended protocols against PostgreSQL and sharded/unsharded PgDog.
• Proposed a prepared EXPLAIN fix preserving parameter types and bindings through generated placeholders; tested reordered parameters and repeated execution on single-database and sharded endpoints.

Merged PR #1778 · EXPLAIN issue & patch #1771

PostHog — Open Source

Oct. 2026 · TypeScript, OpenTelemetry

• Restored Pydantic AI tool inputs and outputs in PostHog’s Conversation view with a TypeScript ingestion patch mapping OpenTelemetry attributes with independent legacy fallbacks.
• Added regression tests for attribute precedence and scalars; validated Pydantic AI instrumentation v3/v5 through Kafka, Node.js, and ClickHouse, plus Playwright UI checks.

Open PR #114989

Semantic Search Engine

Sept. 2026 – Present · Python, TypeScript, MLX, SQLite

• Built local semantic code search for Python, JavaScript, and TypeScript using Qwen embeddings, SQLite keyword/vector retrieval, and incremental indexing with persistent embedding reuse.
• Released a VS Code preview and MCP integration supporting file watching, unsaved-buffer search, cancellation, and token-budgeted context for coding agents.
• Evaluated 106 frozen queries across 5 repositories, retrieving complete source evidence for 83/88 positive cases; measured 175 ms p95 HTTP latency on a held-out repository.

Code

AI-Powered Music Recommendation System

Jan. 2026 – Mar. 2026 · Python, TensorFlow, NumPy, pandas

Infers emotion from facial expressions (FER-2013) and recommends mood-aligned tracks through the Spotify Web API. I trained and fine-tuned a CNN using augmentation and transfer learning to 67.3% accuracy on the held-out test split, running inference on live webcam input.

Code

Electriumap — Electrium Mobility

Dec. 2025 – Present · TypeScript, React, Firebase, Mapbox

A crowdsourced web app that helps riders find EV charging stations, now mapping 2,000+ stations for 75 active users. I maintain the Firebase backend handling geolocation uploads, image storage, CSV imports, and a multi-step form.

Code

Concrete Canoe Programming Team — UWaterloo

Sept. 2025 – Present · Python, Jupyter, TypeScript

Reproducible structural analysis pipelines that model canoe stability and performance. I also maintain the team website and tooling, and integrate analysis output into the GitHub workflow.

Code · Team site

Pathfinder Visualizer

Sept. 2025 – Nov. 2025 · Python, React, JavaScript

An interactive visualizer for shortest-path algorithms — BFS, DFS, Dijkstra’s, A*, and Greedy Best-First — on a grid you can draw obstacles onto and watch the search run.

Code

Chess Opening Trainer

2025 · Python

A command-line opening trainer with a visual board. Openings are defined in JSON, moves are validated as you play, and the engine answers with weighted random responses.

Code

To-Do App

2025 · Python, Flask, SQL, JavaScript

A full-stack task manager with a Flask REST API and MySQL storage — prioritization, category filtering, search, and date-based views, with 87% test coverage.

Code

University of Waterloo

Sept. 2025 – Present · Waterloo, ON

• Bachelor of Software Engineering, cumulative average 87%.
• Varsity Nordic Skiing Team captain.
• Athletics Class Representative, Software Engineering 2030.
• Alumni Women in Technology Scholarship, President’s Scholarship of Distinction, and University of Waterloo Athletic Financial Award.
• Coursework: Software Engineering Methods (Python), Data Abstraction (C++), Programming Principles (C), Sequential Programs, Logic and Computation, Calculus 1–3.