CURRENT · MAR 2026–PRESENT
AI Engineer
Synoverge Technologies
DEVENDRA PARIHAR · AI ENGINEER
Intelligence,
built for the
real world.*
I turn complex AI into useful products. From autonomous agents to scalable ML pipelines, I build systems that work beyond the demo.
4+ years
View my full experience Total experience in AI & data scienceIncluding internships · career started March 2022
A LITTLE HUMAN BEHIND THE AI01 / DEV
Engineer. Builder. Always curious.
From an idea
to something that works.
to something that works.
Where I’ve built.
Full career timeline ↗PREVIOUS · SEP 2024–MAR 2026
AI Developer
Helios Solutions / Sol BI
PREVIOUS · NOV 2022–SEP 2024
Data Scientist
KainSkep
GENERATIVE AI✳AGENTIC SYSTEMS✳SEARCH & RETRIEVAL✳ML ENGINEERING
01 / SELECTED WORK
Built with purpose.
Made to be used.
A selection of experiments, open-source tools, and applied AI projects.
Search & RAG
SemanticLens - Image Search Engine
A semantic image search engine built with Streamlit and Ultralytics, powered by OpenAI's CLIP and Meta's FAISS for intelligent visual content discovery.
Search & RAG
ChromaDB vs PGVector Benchmark Suite
Comprehensive benchmark suite comparing ChromaDB and PGVector performance across insertion, query, memory usage, and CPU utilization scenarios.
Agents & LLMs
AI Calling Agent System
AI-powered calling agent system with advanced voice processing and worklog automation capabilities using Twilio integration.
Search & RAG
Multi Agent Document QA System
A production-ready document Q&A system using cooperating AI agents, Gemini AI, LangGraph orchestration, and Typesense hybrid search with persistent memory and intelligent context engineering.
Developer tools
Jira MCP Server
Comprehensive Python-based MCP server for Atlassian Jira enabling AI assistants like Claude to interact with Jira seamlessly with enterprise-grade features.
Agents & LLMs
Advanced Long-Term Memory System for LLM Agents
Advanced long-term memory system for LLM agents with sophisticated memory architecture and retrieval mechanisms.
Developer tools
Worklog AI for VS Code
VS Code extension that automatically generates detailed worklogs from code changes using AI, available on VS Code Marketplace.
Developer tools
Less Intelligent Vacuum Cleaner
Interactive simulation of a vacuum cleaner with pathfinding algorithms and real-time visualization of cleaning patterns.
Search & RAG
Reasoning Enhanced RAG Chatbot
A chatbot that combines retrieval-augmented generation with chain-of-thought reasoning using FAISS vector search and Flan-T5-Base for explainable answers.
Agents & LLMs
Voice Chat Agent
Interactive Streamlit application with two-way voice communication, leveraging GPT-4 for intelligent responses and edge-tts for text-to-speech capabilities.
Agents & LLMs
Gemini NexusCraft
Streamlit applications leveraging Google's Gemini AI model for powerful text and image generation with configurable parameters and interactive interface.
02 / REAL-WORLD IMPACT
Good engineering.
Measurable difference.
A closer look at the problems, the decisions, and what changed.
90%↗
faster user search
Reported project outcomeCASE STUDY / 01
Making user search feel instant.
The client application was experiencing slow and inconsistent user search performance.
Behind the result
Reason
The slowdown was caused by complex MongoDB queries on a large dataset (20K+ users), especially regex-based searches.
Solution
I replaced the existing search logic with Elasticsearch and built a real-time user sync process that updates Elasticsearch on every create or update operation.
Result
Search performance improved significantly, with response times dropping to under 200 ms; about 90% faster than before.
View benchmark evidence ↗74%↗
faster insert operations
Reported project outcomeCASE STUDY / 02
A leaner foundation for RAG.
The existing RAG application used ChromaDB for vector search and document processing, but ingestion was slow, queries became unpredictable, and ChromaDB often got stuck due to system and memory limits.
Behind the result
Reason
ChromaDB required additional system resources, and its ingestion pipeline was heavy. Since the system was already using PostgreSQL, maintaining a separate vector DB increased overhead and contributed to performance bottlenecks.
Solution
I performed a full benchmark comparing ChromaDB with PGVector to determine the best option for production. Since PostgreSQL was already part of the core architecture, PGVector was the first candidate. The benchmark revealed that PGVector was significantly faster at insertion (5s vs 19.22s), far lighter on memory (32MB vs 66MB), and consumed much less CPU (~4% vs ~18%). Although ChromaDB had faster query performance (0.47s vs 2.77s), PGVector offered better overall efficiency and tunability(this complete experiment was on 10k data rows). Based on these findings, I replaced ChromaDB with PGVector to simplify the architecture and improve system performance.
Result
The migration made the RAG pipeline more stable and scalable. Insert operations became ~74% faster, memory usage dropped by ~50%, and CPU load decreased by ~76%. Infrastructure overhead reduced because vector search and primary data storage now operate within a single PostgreSQL system. Query speed remained strong and is tunable with indexing strategies like ivfflat. Overall, the system became faster, lighter, more reliable, and better suited for production workloads.
View benchmark evidence ↗03 / THE TOOLKIT
The right tools.
For the real problem.
From the first data pipeline to the final deployment, I work across the AI stack.
Have a challenge in mind? ↗01
AI & machine learning
Generative AI · LLMs · NLP · Computer vision
PyTorch · TensorFlow · Scikit-learn
02
Agents & retrieval
LangChain · LlamaIndex · LangGraph
FAISS · PostgreSQL · Typesense · MCP
03
Data & development
Python · SQL · R · PySpark
Pandas · NumPy · Git
04
Cloud & deployment
Microsoft Azure · Amazon Web Services
Docker · ML pipelines
04 / PROFESSIONAL EXPERIENCE
My experience.
From then to now.
4+ years across AI engineering, data science, and early internships. Every role, company, and milestone below.
EDUCATION / 2020–2024
B.Tech in AI & Data Science
Govt. Engineering College Bikaner
March 2026 - Present
AI Engineer
Synoverge Technologies
Building AI-driven products and production-ready intelligent systems for real-world business workflows.
September 2024 - March 2026
AI Developer
Helios Solutions - Sol BI
Architected the "Actionable Business Insights" platform using state-of-the-art LLMs and predictive analytics.
November 2022 - September 2024
Data Scientist
KainSkep
Led GenAI initiatives: Custom Object Detection on Azure, Airflow-based ETL ("PlugData"), and large-scale Databricks pipelines.
July 2022 - September 2022
Machine Learning Intern
OriginCoreTech
Developed core ML models for predictive maintenance and automated classification systems.
May 2022 - June 2022
AI Developer Intern
Learn and Build
Developed AI chatbots and predictive models using ML techniques. contributed to Python backend development.
March 2022 - April 2022
Machine Learning Intern
Suvidha Foundation
Led a team of 20 interns to develop a comprehensive Python course. built automation tools for document processing.
05 / RESEARCH
Curiosity,
put to paper.
Exploring satellite imagery analysis through Vision Transformers. Published in IEEE InGARSS 2024.
06 / NOTES FROM THE WORKBENCH
Things I’ve learned.
Ideas worth sharing.
07 / LET’S CONNECT
Have something
in mind?
A challenging problem, an interesting idea, or just a conversation about AI. I’d love to hear it.
devendraparihar340@gmail.com ↗