AMEYA KASHID

DATA ENGINEER + APPLIED AI ENGINEER

Based in Tampa. Explore my projects, experience, and credentials through the terminal.

ameya@ledger: ~/career.db
Type read-only SQL against the career ledger, or use the suggested queries below.
Black-and-white manga ink portrait of Ameya KashidON DEMAND
AMEYA KASHID — INK ILLUSTRATION
Black-and-white manga ink illustration: a lone figure seen from behind facing a vast canyon of server racks with rivers of data flowing between them
THE DATA WORLD — FIELD SKETCH, INK ON PAPER

第02話 MANIFESTO

I move data for a living. SQL, Python, Kimball-style dimensional models, and cloud platforms end to end — ingestion, transformation, modeling, BI.

I also build AI workflows on those systems, including OCR, document intelligence, computer vision, and LLM-backed workflows with guardrails.

Use the terminal to explore my career data and the case studies to read about individual projects.

PERSONAL UPTIME — SINCE 1995-11-01

NO DOWNTIME RECORDED.

MS IS&T — DATA SCIENCE & ANALYTICS · CGU · GPA 3.925

EC-COUNCIL NDE · NETWORK SECURITY → INCIDENT RESPONSE

第03話 — THE PIPELINE

FOUR STAGES. ONE HABIT.

STAGE 01

INGEST

Every system starts as a flood. Raw documents, events, tables — somebody has to get them in.

  • AZURE DATA LAKE
  • AWS S3
  • SQL SERVER
  • POSTGRESQL
  • MONGODB

STAGE 02

TRANSFORM

Raw is useless. This is the part nobody sees and everybody depends on.

  • AZURE SYNAPSE
  • SNOWFLAKE
  • DBT
  • KIMBALL
  • PYTHON
  • SQL
  • JAVA

STAGE 03

MODEL

Star schemas, detections, fine-tunes. Models are just tables that learned something.

  • YOLOV8
  • TESSERACT OCR
  • GEMINI
  • LLAMA
  • DEEPSEEK
  • LORA
  • KAFKA
  • SPARK STREAMING

STAGE 04

SERVE

Data earns its keep when someone makes a decision with it.

  • TABLEAU
  • POWER BI
  • REGRESSION / TIME SERIES
  • MONTE CARLO
  • EXCEL
  • SAS

BIZIZZ 2020 → KATE INDUSTRIES 2021 → CGU 2023 → LID VIZION 2024 —

第04話 — THE WORKS

SELECTED WORK

Explore the case studies.

CH.01APPLIED AI LIVE
Black-and-white manga ink illustration: an engineer at a terminal while a spectral ink fox-spirit made of flowing data streams rises from the glowing screen

Ask the Warehouse

Production text-to-SQL assistant analysts actually use: semantic-layer allowlist, self-correction, eval-gated deploys — every answer shows its SQL.

LangGraphMCPText-to-SQLdbt SemanticEvals

READ THE CHAPTER →
CH.02STREAMING
Black-and-white manga ink illustration: a frozen lake holding table-shaped icebergs beneath an aurora of streaming data ribbons, a lone cabin on the snowy shore

Real-time Streaming Lakehouse

Migrated nightly batch to a Kafka → Flink → Iceberg lakehouse with dbt on streaming tables; dashboards run on the current hour.

KafkaFlinkIcebergdbt

READ THE CHAPTER →
CH.03APPLIED AI LIVE
Black-and-white manga ink illustration: an infinite library where a lone figure pulls luminous threads of light from open books into a single magnifying lens

Document RAG with Measured Retrieval

Document RAG reviewers trust because every answer is cited and retrieval quality is a measured gate, not a vibe — exposed to agents over MCP.

EmbeddingsVector SearchRerankingEvalsMCP

READ THE CHAPTER →
CH.04LLMS
Black-and-white manga ink illustration: a grand laboratory balance scale weighing a speech bubble on one pan against a golden reference tablet on the other

LLM Evaluation Harness

Golden sets, CI regression gates, and observability that turned LLM releases from leaps of faith into gated deploys.

Golden SetsCI GatesObservabilityPrompt Versioning

READ THE CHAPTER →
CH.05DATA ENG
Black-and-white manga ink illustration: two monumental database towers joined across a chasm by a bridge of flowing luminous packets, messenger birds flying one direction

Change Data Capture, Exactly Once

Replaced nightly full-table syncs with Debezium → Kafka → Iceberg upserts; idempotent consumers, managed schema evolution, replayable history.

DebeziumKafkaIcebergSchema Registry

READ THE CHAPTER →
CH.06DATA ENG
Black-and-white manga ink illustration: a grand torii gate standing in a river of flowing data packets, an abacus and coin scales on the bank, a guardian with a ledger

Data Reliability & FinOps

Moved breakage discovery from board meetings to CI: data contracts, dbt tests, anomaly baselines, per-model cost attribution.

Data Contractsdbt TestsAnomaly DetectionFinOps

READ THE CHAPTER →
CH.07GIS
Black-and-white manga ink illustration: rolling California hills in topographic ink lines, wildfire smoke in dense screentone beyond the ridge, a surveyor with a map, a satellite overhead

GeoAI Wildfire Risk, California

Spatial ML over terrain, fuels, and weather rasters — California wildfire risk as a living map, not another spreadsheet.

Geospatial MLSatellite ImageryRasterPython

READ THE CHAPTER →
CH.08APPLIED AI
Black-and-white manga ink illustration: a great mechanical eye on an articulated arm inspecting floating documents, glowing glyphs lifting off the paper as light

Multimodal Document Intelligence

Production extraction at Lid Vizion: CV preprocessing + OCR + LoRA-tuned VLM, with a confidence-ranked queue so reviewers judge exceptions, not key fields.

OCRVLMLoRAHuman-in-the-Loop

READ THE CHAPTER →

第06話 — THE CONSTELLATION

ONE LEDGER, DRAWN.

EVERY EDGE IS A ROW IN THE SAME DATABASE.

HOVER A NODE — THE MAP IS A QUERY.

CH.01CH.02CH.03CH.04CH.05CH.06CH.07CH.08SYNAPSESNOWFLAKEDBTKIMBALLPYTHONSQLJAVATABLEAUPBIREG/TSMONTE CARLOEXCELSASARCGISSPATIALYOLOV8TESSERACTGEMINILLAMADEEPSEEKT5LORAKAFKASPARKADLS3MSSQLPOSTGRESMONGONETSECTHREATIDSRISKAPACHE ICEAPACHE FLIDEBEZIUM CSCHEMA REGDATA CONTRANOMALY DEFINOPS / CLANGGRAPH MCPTEXT-TO-SQLLM EVALUAVECTOR SEARERANKINGLLM OBSERVVLM (VISIOHUMAN-IN-TGEOSPATIALSATELLITE

第07話 — SIGN OFF

CONTACT ME

Write to me about data engineering, applied AI, or dimensional modeling.

ameya.kashid@cgu.eduPRIMARY
ameya2611@gmail.com

GITHUB ↗LINKEDIN ↗

HOVER TO INK THE SEAL