Kishan K
Blog

Engineering notes from production AI.

Honest write-ups on fine-tuning, evaluation, computer vision and LLM systems: what I measured, what I got wrong, and what I would do differently.

A 5-line baseline humbled my 99.6%

I fine-tuned a 7B model to 270/271 on a held-out set. Then I trained TF-IDF + logistic regression on the same data and got 269/271. Here is what I changed in how I report results.

EvaluationFine-tuningLLMs
2026-11-03 · 2 min read

Tracking microscopic swimmers with YOLOv8 + DeepSORT on 2 vCPUs

How a microscopy video becomes a motility report: detection, tracking, a queue, and where the time actually goes.

Computer VisionTrackingBackend
2026-11-03 · 2 min read

How to backtest a price forecaster without fooling yourself

Your service's own error metric is probably measured on training data. Here is a rolling-origin backtest recipe, with a naive baseline in the room.

ForecastingTime SeriesEvaluation
2026-11-03 · 2 min read

Cache your LLM like a database: bucket the inputs, template the output

When thousands of users ask almost the same question, don't pay for thousands of generations. A two-level template cache pattern.

LLMsCachingSystem Design
2026-11-03 · 2 min read

When your validation set is secretly your training set

Fine-tuning on templated data makes leakage easy to create and easy to miss. The checks I run now, and how I report results when the leakage cannot be removed.

Fine-tuningEvaluationData
2026-11-03 · 2 min read

How I test an AI agent that must never invent a source

Sentinel drafts equity research with page-level citations, then audits itself for SEBI rules. Here is how I tested it, what I measured, and the number I stopped quoting.

AgentsEvaluationTesting
2026-11-04 · 4 min read