Software Engineer · AI Systems & Data Engineering

Sai Hemanth
Mandala

Building RAG & LLM systems and high-throughput data pipelines — from raw streams to answers humans trust.

hemanth@portfolio:~$
Retrieval-Augmented GenerationApache SparkKafka StreamingVector SearchChange Data CaptureLLM PipelinesAWS Retrieval-Augmented GenerationApache SparkKafka StreamingVector SearchChange Data CaptureLLM PipelinesAWS
01 — Profile

Profile

Software engineer with 1.7 years turning messy data into intelligent, real-time systems.

I work where AI meets data infrastructure — designing RAG/LLM applications that answer from documents, and the streaming pipelines that feed them. My focus is the unglamorous engineering that makes intelligence reliable: profiling, lineage, quality checks, and throughput.

Core stack runs Python, Apache Spark, Kafka, Node.js and AWS — with a habit of shipping things that measurably move the needle.

~50%Faster data-profiling execution
30–40%Less data-investigation time
1.7yProduction engineering
Pipelines kept flowing
02 — Experience

Experience

Jan 2024 — PresentRemote

Software Engineer

DHIRA — Digital Human Interface & Robotic Applications
  • Built a context-aware RAG Q&A system over textbooks using Azure Translation, Chroma DB & GPT-4o for enriched prompts.
  • Refactored data-profiling logic, lifting execution speed ~50%.
  • Shipped an AI resume-to-JD matching tool with Azure Document AI + GPT-4o and a Gradio match-score UI.
  • Automated media generation from PDF extracts — image scenes via Azure OpenAI & DALL·E, video via Luma AI.
  • Built real-time DB replication: MariaDB → PostgreSQL with Debezium & Kafka, processed via Spark Streaming.
  • Built a Delta Lake pipeline querying delta tables across Spark, Hive & HDFS.
  • Wrote ETL transformations — column splitting, date rearrangement, Box-Cox refinement & quality checks.
  • Scripted entity registration in Apache Atlas, cutting investigation time ~30–40%.
  • Enhanced an NLP-to-SQL workflow so non-technical users query databases directly.
Jun 2023 — Sep 2023Remote

Data Science & ML Intern

Gilbert Research Centre
  • Applied ML techniques incl. linear regression inside a research team.
  • Handled preprocessing, feature engineering & model testing; exposure to medical instruments.
Aug 2019 — Jul 2023Tekkali

B.Tech — Computer Science

Aditya Institute of Technology and Management · CGPA 7.78 / 10
03 — Selected Builds

Selected Builds

PRJ_01

Hemanthify

A social platform with real-time chat, post sharing, likes, comments and follows — Socket.io chat engine, Google OAuth 2.0 login & forgot-password flow, deployed on AWS EC2 (t2.micro, Ubuntu).

Node.jsMongoDBSocket.ioOAuth 2.0AWS EC2
PRJ_02

RRR — Reviews & Ratings

A restaurant platform: owners manage menus & profiles, reviewers post star ratings and item-level reviews. Google OAuth 2.0, deployed on AWS EC2. Published as a peer case study.

Node.jsExpressMongoDBOAuth 2.0Published
04 — The Stack

The Stack

AI / ML // reasoning

RAGLLMsGPT-4oPrompt EngineeringChroma DBVector SearchAzure Document AIDALL·ELuma AINLPGradio

Data Engineering // throughput

Apache SparkPySparkKafkaSpark StreamingCDCDebeziumHiveDelta LakeHDFS

Languages // foundations

PythonJavaCJavaScriptShell

Backend & Web // services

Node.jsExpress.jsREST APIsSocket.ioEJSjQueryAjaxSASSReact

Data Quality & Governance // trust

ETLData ProfilingQuality ChecksApache AtlasData Lineage

Databases & DevOps // ground

MongoDBPostgreSQLMariaDBLinuxGitNginxAWS EC2
05 — Certifications

Certifications

Coding Ninjas Front End — Full Stack Web Development View certificate ↗ Coding Ninjas Back End — Full Stack in Node.js View certificate ↗ { } Coding Ninjas Introduction to Java View certificate ↗ >_ Coding Ninjas Data Structures & Algorithms in Python View certificate ↗ Recognition Letter of Recommendation View letter ↗ IJRACSE · Research Published Case Study Read publication ↗
06 — Published

Published

IJRACSE · Case Study

Restaurant Reviews & Ratings — System Design & Results

A peer case study covering the RRR platform's architecture, step-by-step implementation and measured outcomes — documentation, technical depth and problem-solving in one record.

Read Publication
ChessRubik's CubeOrigamiPhotoshopTech-Exploration ChessRubik's CubeOrigamiPhotoshopTech-Exploration
07 — Let's build

Got data that needs intelligence?
Let's talk →

m.saihemanth1@gmail.com