Mallikarjun Reddy

GenAI / AI Engineer · Bengaluru

Summary

GenAI and Systems Engineer specializing in orchestrating Large Language Models (LLMs), deterministic state graphs via LangGraph, and scalable multi-modal AI architectures. Proven track record of architecture and end-to-end deployment of enterprise-grade RAG pipelines, fine-tuning state-of-the-art models (Claude, Llama) using LoRA/PEFT, and building highly-concurrent ML microservices. Adept at bridging bleeding-edge AI research with production-grade engineering using Claude Code for agentic pair programming to build programmatic systems.

Education

Indian Institute of Technology (IIT) Roorkee

CGPA 7.4/10 · JEE Advanced rank 3015

Experience

Kredily — Senior GenAI & AI Systems Engineer

• Built GenAI and RAG workflows for HR and payroll intelligence, enabling natural-language queries and automated reporting.
• Improved services handling roughly 20M requests/day, reducing evening peak compute demand by about 40% through caching and read-query optimization.
• Rolled out Attendance Regularisation V2 and the Payroll Arrears engine to all customers; strengthened Celery task and cloud workload reliability.

Isourse Technologies — AI/ML Engineer

• LLM fine-tuning & deployment: Finetuned and deployed state-of-the-art open-source LLMs (Llama 3, Qwen, DeepSeek) using SFT, PEFT, and LoRA to architect a robust Text-to-SQL conversational BI platform (using Pydantic), completely automating enterprise dashboard generation.
• Agentic RAG systems: Developed a scalable RAG-based AI chatbot orchestrated with persistent agentic memory and vector stores (Milvus/pgvector), drastically improving contextual accuracy and driving a massive boost in user retention.
• AI-powered ATS: Engineered an Intelligent ATS system leveraging custom vector embeddings and semantic similarity, automating candidate evaluation and achieving near-human logical scoring accuracy at high latency efficiency.
• High-volume data engineering: Built and orchestrated a fault-tolerant, real-time data warehousing ETL pipeline (APIs → Kafka → PySpark) capable of seamlessly processing ultra-high-volume distributed data streams.
• MLOps & infrastructure: Spearheaded the migration of legacy ML infrastructure to highly concurrent, Dockerized FastAPI microservices, implementing cluster observability and alerting via Grafana and Prometheus.

Relevant projects

ZENVidio

• End-to-end multi-agent architecture: Architected an autonomous video platform via LangGraph state graphs and MCP tool calls to generate Remotion code.
• Vision AI feedback loop & scalable render engine: Developed an “Agentic Mastery Loop” using PIL/Vision AI to auto-critique visual layouts and autonomously rewrite .tsx files, concurrently bridging Python AI backends with a Chrome/Puppeteer pipeline.

3D Avatar Motion Capture

• Browser-based inference: Architected an ultra-low latency, real-time 3D avatar system using MediaPipe for perfect alignment of full-body, hand, and complex facial mappings strictly via webcams.

Enterprise GenAI Chatbot

• Multi-modal discovery: Spearheaded a corporate AI assistant leveraging hybrid search (semantic + exact match) with Milvus and real-time bi-directional FastRTC voice capabilities to accelerate big-data extraction.

Technical skills

Generative AI & ML: LLMs (Claude/GPT/Llama), LangGraph, Agentic Workflows, Text-to-SQL, Multi-Modal AI, LoRA, Vector Search (Milvus/pgvector)
Languages & frameworks: Python, SQL, FastAPI, Pydantic, PyTorch, TensorFlow, LangChain, LlamaIndex, Node.js, React
Data & cloud: Kafka, PySpark, PostgreSQL, Snowflake, Azure AI Foundry, GCP, AWS, Docker, Kubernetes
Developer tools: Claude Code, Model Context Protocol (MCP), Git, CI/CD, ONNX, Grafana, Prometheus