Foundry Capital
Boutique broker-dealer · ~$400M AUM
Replaced an Excel macro with a signed-manifest reconciliation in 3 weeks
Stand up a reliable broker integration with nightly reconciliation in three weeks.
A small fintech needed to connect their internal book to a clearing broker's API and reconcile positions, cash, and corporate actions every night without manual intervention. Their existing approach was a CSV download + Excel macro maintained by one person.
Throughput
584/s
p95
32ms
Errors
0.02%
- Industry
- Fintech
- Timeline
- 3 weeks
- Team
- 2
- Service
- Software
- Project tier
- Custom Software / $14,995
The Problem
What was broken.
Trades reconciled by hand against a broker CSV, every morning, by one person. When she was on PTO, reconciliation didn't happen. The few discrepancies that slipped through compounded into reporting errors that took half a day to unwind. Auditors had started to ask sharper questions about the manual control.
Our Approach
How we framed it.
Treated reconciliation as a workflow problem, not a script. Used Temporal to orchestrate the nightly job: pull broker positions, pull internal book, diff, classify discrepancies (timing vs. real break), file resolution tickets in Linear automatically. Kept the existing CSV pipeline alive in parallel for 30 days as a safety net — diffed both outputs and alerted only when they disagreed.
Capability proof
What this case demonstrates.
This case makes the hidden work visible: strategy, architecture, delivery control, quality evidence, and handoff.
01 / Product judgment
Problem framed before UI
Trades reconciled by hand against a broker CSV, every morning, by one person. When she was on PTO, reconciliation didn't happen. The few discrepancies that slipped through compounded into reporting errors that took half a day to unwind. Auditors had started to ask sharper questions about the manual control.
02 / Technical depth
8 stack decisions
TypeScript (Node.js), Postgres, Temporal (workflows), REST + FIX, Sentry, Datadog
03 / Delivery discipline
4 delivery checkpoints
Mapping + sandbox / Diff engine prototype / Temporal workflow + retries
04 / Handoff quality
5 shipped artifacts
Temporal-orchestrated nightly reconciliation / Postgres-backed break ledger with audit trail / Linear integration: real breaks become tickets with prefilled context
Production artifacts
Inspect the work behind the visible result.
Each case exposes the surfaces, systems, evidence, and handoff package that make the shipped product usable after launch.
Experience layer
Buyer or user surface
Temporal-orchestrated nightly reconciliation. Idempotent activities, signed manifest per run, 30-day parallel-run before retiring the legacy macro.
Proof 01
Stand up a reliable broker integration with nightly reconciliation in three weeks.
Proof 02
Got read-only sandbox credentials from the broker. Mapped every field the existing macro touched.
Proof 03
Datadog dashboard + PagerDuty escalation
Production signals
Observable
Errors, logs, alerts, or dashboards included.
Risk-aware
Security and compliance boundaries named.
Handoff-ready
Owner can keep operating after delivery.
Before / after · product UI mockup
Industry · Fintech
Before:Reconciliation lived in 3 brittle Excel sheets; broker fills were copy-pasted nightly.
After:Temporal-orchestrated nightly reconciliation. Idempotent activities, signed manifest per run, 30-day parallel-run before retiring the legacy macro.
How the engagement ran.
- 01Day 1-2
Mapping + sandbox
Got read-only sandbox credentials from the broker. Mapped every field the existing macro touched.
- 02Day 3-5
Diff engine prototype
Built the diff classifier locally against three weeks of historical CSV + sandbox data.
- 03Week 2
Temporal workflow + retries
Wrapped the diff in a Temporal workflow with idempotent activities, exponential backoff, and a manual approval gate for high-value breaks.
- 04Week 3
Parallel-run + handoff
Ran new pipeline alongside the legacy macro for the last week. Wrote the runbook and trained two people.
- 1
Day 1-2
Mapping + sandbox
Got read-only sandbox credentials from the broker. Mapped every field the existing macro touched.
- 2
Day 3-5
Diff engine prototype
Built the diff classifier locally against three weeks of historical CSV + sandbox data.
- 3
Week 2
Temporal workflow + retries
Wrapped the diff in a Temporal workflow with idempotent activities, exponential backoff, and a manual approval gate for high-value breaks.
- 4
Week 3
Parallel-run + handoff
Ran new pipeline alongside the legacy macro for the last week. Wrote the runbook and trained two people.
Deliverables
What we shipped.
- ✓Temporal-orchestrated nightly reconciliation
- ✓Postgres-backed break ledger with audit trail
- ✓Linear integration: real breaks become tickets with prefilled context
- ✓Datadog dashboard + PagerDuty escalation
- ✓Runbook covering common break types and resolution steps
Outcomes.
engagement targetsGoal: zero-touch nightly reconciliation across positions + cash
Goal: discrepancies surface as classified tickets, not silent failures
Plan: 30 days of parallel-run before retiring the legacy macro
Plan: audit-friendly: every reconciliation run produces a signed manifest
Plan: on-call runbook with paging only on real breaks, not timing diffs
Honest challenges
What we got wrong (or almost wrong).
The pretty version of any case study skips this part. We don't.
- 01
Broker's sandbox returned slightly stale data; we wrote a tolerance window for end-of-day timing differences and documented it.
- 02
Corporate actions (splits, dividends) needed special handling — added a per-action classifier with a human-review queue.
- 03
Temporal's local dev environment had a quirk on Windows; pinned to Docker Desktop with a short README to save the next engineer the same hour we lost.
In our own words
Reconciliation isn't a script, it's a workflow. Modelling it in Temporal with idempotent activities and a 30-day parallel run meant the cutover was uneventful — which, in money plumbing, is the highest praise the work can earn.
From the Hayaiti team
Engineering · design · security
Technical blueprint
How the work holds together.
Buyers should see that the visual layer is backed by architecture, quality gates, and operational ownership.
Experience
1Application
2Data
3Operations
4Security
5Stack used
8 technologiesRelated
Other cases like this.
Northwind Studios
Onboarding rebuild · Series A SaaS
Cut signup-to-aha-moment from 9 minutes to under 90 seconds.
Quant TradingHayaiti Research
Cross-sectional momentum harness
Stand up a reproducible backtest harness for a momentum strategy on equities.
Climate / ESGSentinel Operating
Scope 1-2-3 emissions dashboard
Produce defensible Scope 1-2-3 emissions reports without an army of consultants.
Want a case study like this?
Want this level of production quality on your project?
Send a short brief and we'll reply with scope, timeline, price direction, and the first technical recommendation.