Burak Tekin—

Software engineer · Backend & systems

Burak Tekin

I build reliable backend systems, real-time and data-intensive software, and automation that removes fragile manual work.

View selected work

Engineering with explicit constraints, observable behavior, and evidence before claims.

02

Selected work

One focused case study. More should earn their place.

Featured case study

Ongoing

PolyGo

Execution timing changed the distribution more than the headline PnL suggests.

Historical replay — not live returns.

Role
Backend and systems engineering
Technology
Java · Python · WebSockets · Docker

Problem

Headline backtest net can hide how latency, fees, no-fills, and tail outcomes change the distribution underneath it.

Approach

Compare the same paired historical cohort across explicit exit-latency scenarios, then inspect selected replays and unresolved execution states separately.

Outcome

The evidence supported a clearer execution-risk model while remaining insufficient for production deployment.
775Paired historical trades
~165 hours6.9 days observed
500 / 1000 msExit-latency stress tests
1No-fill exposure

Execution timing study

Historical replay

Loading public release…

Evidence & decision

What the evidence supports.

Research prototype—not a deployed income system.

Median latency impact
$0

Tail range −$39.38 to +$45.10

Fee drag at 1000 ms
51.6%

$9.16k gross · $4.73k fees

Unresolved no-fill
1 / 775

175 shares · $99.75 entry notional

Decision factors
  • The dataset was repeatedly inspected; this is not an untouched holdout.
  • Execution timing is non-monotonic: the median impact is $0 while tail outcomes move the aggregate.
  • One no-fill remains unresolved, and fees consume roughly half of gross at 1000 ms.
Methodology details
Fair paired closed cohort baseline
$3,989.97109

774 filled

500 ms closed-only net
$3,939.91579
1000 ms closed-only net
$4,429.33679
1000 ms execution
774 filled · 1 no-fill

51.03% win rate

Median latency impact
$0

Worst −$39.38459 · best +$45.09924

1000 ms gross / fees
$9,160.60480 / $4,731.26801

Fees consume approximately 51.6% of gross

03

Capabilities

Practical strengths for systems that have to keep working.

01

Backend engineering

Clear service boundaries, explicit contracts, and maintainable domain logic.

02

Reliability

Failure-aware flows, bounded retries, and behavior that remains inspectable under pressure.

03

Observability

Useful logs, traceable operations, and signals designed for the person diagnosing the system.

04

Data pipelines

Chronological processing, replayable inputs, and careful separation of source data from inference.

05

Automation

Pragmatic tooling that replaces repeated manual work without hiding important decisions.

04

About

I prefer small, explicit systems over clever ones. My approach is to understand the real constraint, make behavior observable, and verify the result before calling it done.