10+ PROJECTS • COMPLETED • SINCE 2023 •
★
COMPLETED
Since 2023

Hi! I Am

Sahojit Karmakar.

Sahojit
AI/ML
View Resume
ENGINEERARCHITECT

& Data Systems Builder

AI/ML Engineer & Data Systems Builder.

Building as an AI/ML Engineer & Data Systems Architect with:
PythonPyTorchAWS BedrockApache SparkKafkaAirflowLangChainOpenSearchDockerKubernetesdbtPostgreSQLTensorFlowRustPythonPyTorchAWS BedrockApache SparkKafkaAirflowLangChainOpenSearchDockerKubernetesdbtPostgreSQLTensorFlowRustPythonPyTorchAWS BedrockApache SparkKafkaAirflowLangChainOpenSearchDockerKubernetesdbtPostgreSQLTensorFlowRustPythonPyTorchAWS BedrockApache SparkKafkaAirflowLangChainOpenSearchDockerKubernetesdbtPostgreSQLTensorFlowRust
WHO I AM

About

Hi, I am Sahojit Sahojit Karmakar , practicing ML and Data Engineering since 2023, focused on building intelligent systems , data pipelines, and LLM products.

Open to AI/ML & Data Engineering roles ★

Expertise in Tools

Python
Rust
PyTorch
n8n
AWS
Claude
Kafka
LangChain
Docker
PostgreSQL
Git
more
WHAT I OFFER

Expertise

EXPLORE

01

ML Engineering

Designing and deploying production ML models — training pipelines, model serving, A/B testing, and monitoring.

EXPLORE

02

Data Engineering

Building robust ETL/ELT pipelines, data lakes, and real-time streaming systems using Spark, Kafka, and Airflow.

EXPLORE

03

LLMOps & RAG

Architecting RAG pipelines, vector search systems, and LLM evaluation frameworks on AWS and open-source stacks.

EXPLORE

04

MLOps & Cloud

End-to-end MLOps on AWS — SageMaker, Bedrock, Lambda, OpenSearch — with CI/CD, cost monitoring, and observability.

My Portfolio

Featured Works

A collection of ML Engineering, Data Pipeline Design, LLMOps, and AI Systems projects.

2024

AIOPSPIPELINE RCA

CASE STUDY

ML ENGINEERINGMLOPS
2024

AUTONOML

CASE STUDY

AUTOMLDATA ENGINEERING
2025

MODEL ARBITRATION ENGINE

CASE STUDY

LLMOPSSYSTEM DESIGN

Thoughts & Guides

Blogs

AI/MLRAG

Building a Production RAG Pipeline: What I Learned the Hard Way

From chunking strategies to re-ranking and observability — the things nobody tells you when you move a RAG system from demo to production.

July 2025 · 8 min read→
LLMOpsSystem Design

How I Cut LLM Inference Cost by 84% with Epsilon-Greedy Routing

A deep dive into the Model Arbitration Engine I built on AWS — bandit algorithms, EWMA latency tracking, and the math behind intelligent model selection.

June 2025 · 6 min read→
Data Engineering

Silent Data Loss: The ETL Bug That Doesn't Crash Anything

How rows can silently vanish in your pipeline without a single error log — and how to catch it using statistical process control before your analytics rot.

May 2025 · 5 min read→

My Journey

Experience

Futurense Technologies

Current Employer

Building multi-agent LLM systems (research, writer, and critic agents) with real LLM calls, web research, financial data tooling, and persistent memory. Instrumenting pipelines end-to-end with LangFuse tracing and eval hooks for reproducible agent behaviour.

LLM SystemsApplied GenAILangGraphFastAPILangFuse

Lovely Professional University