# Sprint record, March to October 2026

Anonymized record of a real engagement for a US investment analytics platform, 2026.

The engagement began in August 2025. This record covers the written sprints from March 2026.

The client runs a platform where analysts upload Python scripts. The platform runs the scripts every night against market data. Members read return cards, fund drill downs, PDF reports and a daily email. The stack is Python and Django, React and TypeScript, PostgreSQL, AWS containers and scheduled tasks, Terraform and GitHub Actions.

All work was async. The client wrote a request, a review or a list of bugs. I wrote back a scope, then updates, then pull requests, then a report with evidence.

Every count below comes from the delivered reports, time statements, commit lists and pull request records. A dash means the source folder doesn't hold that count. Names, domains, repository names, billing numbers, rates and amounts are removed.

## The sprints at a glance

| Sprint | Dates (2026) | Days | Theme | Counts from the record | Hours billed |
|---|---|---|---|---|---|
| 1 | Mar 17 to May 8 | about 7 weeks, part time | Platform recovery after a takeover | 5 linked faults resolved, 11 failure points fixed, 23 files changed, 14 resolution items (12 deployed, 1 in progress, 1 client step), coding rules audit of 11 Python files | 25 in the last week. Earlier weeks are not in the folder |
| 2 | May 11 to 15 | 5 | Infrastructure audit, dead task broker removed | 147 managed cloud resources audited, 11 task definitions corrected, 4 client bugs root caused, 17 scripts confirmed running again on day 2 | 14 |
| 3 | May 14 to 15 | 2 | Script runner fix | 301 of 331 scripts failing, traced to shared output buffers between threads and to containers at 512 MB. Success rate from 0.6 percent to 82 percent. 3 commits deployed | 6.5 |
| 4 | May 18 | 1 | Three remaining bugs and a script guide | 3 bugs investigated, 2 ready to deploy the same day, deliverables checked in 5 audit passes | 4 |
| 5 | May 18 to 22 | 5 | Sign in, registration, password reset, report pipeline | Reports producing PDFs from 0 of 9 to 8 of 9 after a 72 day outage. Script success from about 10 to about 95 percent. Stuck runs now recover after 1 hour | 32 |
| 6 | May 25 to 29 | 5 | Security hardening and report pipeline restoration | 18 security and infrastructure changes, 15 files changed, 3 files added, 8 critical findings closed in code | 25 |
| 7 | Jun 1 to 4 | 4 | Legacy external API retired, full script QA | 63 plus 27 plus 28 scripts moved to the platform database, 116 dead settings removed, 342 scripts validated with a screenshot each, 334 succeeding and 8 failing at the end, 15 commits | 40 |
| 8 | Jun 14 to 17 | 4 | Blank charts, flagged scripts, mail cutover | 5 flagged scripts fixed and checked live, a migration boundary fault fixed, backend test gate at 36 tests, 247 scripts scanned | 15 |
| 9 | Jun 23 to 26 | 4 | Static chart default, output fix, API access | Wrong chart output fixed at the source, admin sign in restored, working API access from the client's own PC, 3 written answers | 16.5 |
| 10 | Jul 1 to 5 | 5 | Output guard, price ingestion, report email | 5 of 5 items live. A blank render can no longer pass as a success. 5,773 price rows loaded. Ticker field widened from 5 to 12 characters by migration | 21 |
| 11 | Jul 9 to 16 | 6 | Script source access, report sender, naming rule | 6 workstreams deployed. A claims check of 12 statements found 11 verified and 1 overstatement, which I corrected | 25 |
| 12 | Jul 21 to Aug 3 | 14, two parts | Report email delivery, analysis sandbox, annualized returns | 15 validation runs in a row on the sandbox. Card returns for 1, 2, 5 and 10 years. Data audit of 24 funds, 0 fabricated values, 10 year mean difference 0.035 points | 57.25 |
| 13 | Aug 4 to 16 | 13 | Cards from 4 to 16 funds, server side search and sort | 12 of 12 behaviour checks, 0 TypeScript errors, production release, developer access | 73, plus 10.25 not charged |
| 14 | Aug 19 to 24 | 6 | Script upload and filing blockers, one daily run | 1 pull request, 9 commits, 41 files, 8 blocking faults closed, bulk filing of up to 500 scripts, 11 step daily run replaced 6 schedules | 36.5, plus 4.75 not charged |
| 15A | Sep 2 | 1 | Bounded, restartable price history preload | 2 pull requests, 8 files, 1 migration, 52 backend tests passing | 5 |
| 15 | Sep 4 to 7 | 4 | Six points from a written review | 4 pull requests, 30 commits, 73 files, 1 migration, tests from 52 to 166, a 3 month infrastructure lock cleared | 36, plus 2 not charged |
| 16 | Sep 17 to 23 | 7 | Fund drill downs, one list process | 3 pull requests, 60 commits, 72 files, 3 migrations, tests from 169 to 341 | 42, plus 3.5 not charged |
| 17 | Sep 30 to Oct 1 | 2 | Three urgent analyst requests | 1 pull request, 60 commits, 59 files, 4 migrations, tests from 342 to 439, 11.8 million history rows loaded in 34 minutes | 32.5 |
| 17B | Oct 2 to 7 | 6 | Security, script sandbox, health email | 1 pull request, 100 commits, 189 files, 3 migrations, tests from 439 to 683 | 37, plus 4 not charged |
| 18 | Oct 11 to 19 | 9 | New return window, 1,800 row drill down, dated reruns | 6 pull requests, 77 commits, 116 files, 3 migrations, tests from 683 to 922, 441 report statements checked against the live site | 54 |

## Totals from the record

- 20 sprints in 31 weeks, from March 17 to October 19, 2026.
- 597.25 hours billed where the time statement is in the folder. Another 24.5 hours weren't charged, because they went on merges, reviews and fixes to my own work.
- Sprints 13 to 18 came to 316 billed hours, with 5 of them in sprint 15A. The seven sprints on the sprints page (13 to 18 without 15A) merged 16 pull requests with a record, and their commit lists hold 336 commits. Sprint 15A added 2 more pull requests.
- The backend test suite grew from 52 tests at the start of sprint 15 to 922 at the end of sprint 18.
- After sprint 1, sprint length ran from 1 day to 14 days. The median sprint was 5 days.

## How the engagement changed over time

### March to May, recovery

The platform had stopped refreshing data on March 9. One reported fault turned out to be five linked faults across scheduling, email records, redirects, monitoring and a data scraper. The first sprints were short and diagnostic. Each one closed a root cause and gave the client a plain written account of it.

### June to July, reliability

Work moved to the script fleet. I retired a dead external API that 118 scripts still used, built a QA pass that opened every script page and saved a screenshot, and added guards so a blank chart can never count as a success.

### August to October, product work

With the base stable, sprints became feature sprints with a written scope and hours per item. Return cards, fund drill downs, a script sandbox, a health email and dated reruns. Tests, migrations and review passes grew with them.

## Sources

Sprint reports, time statements, delivery messages, commit lists and pull request records in the client folder, read only. Hours come from the time statements.