Midhilesh Momidi
Senior ML Engineer| Applied AI | AI Engineer | Master @ Codeforces | Ex-Walmart
Bengaluru
10+ years of work experience
Sep 20251 yrPresent
Bengaluru, Karnataka, India
Built a Multi-agentic SRE system on Microsoft Agent Framework with fine-tuned task-specific agents that autonomously triage incidents end-to-end correlating Kusto telemetry and logs, executing remediation runbooks, raising ICMs with structured RCAs, and generating TSGs and bug analysis dashboards for recurring fault codes reducing manual intervention by 60%.
Built Coolville Copilot, an AI assistant enabling firmware failure investigation and automated remediation across 12,500+ nodes daily, with Text-to-KQL translation for large-scale telemetry querying and agent-driven ICM creation for incident tracking.
Developed a Multimodal RAG system over PDFs, PPTs, flowcharts, ADO tasks, and knowledge articles using Azure AI Search with hybrid retrieval and PyTorch-based embedding models for cross-modal feature extraction, mapping firmware fault codes to step-by-step resolution workflows for firmware engineers.
Integrated evaluation frameworks within the CoreIQ AI platform to enable onboarding of AI agents, systematic bench marking and deployment of Microsoft Foundry models.
Nov 202411 mosSep 2025
Bengaluru, Karnataka, India
Developed a Multi-Agentic RAG architecture using LangGraph with Redis for rapid agent-state retrieval and FastRTC for real-time voice (STT, TTS) and chat agents handling order tracking, returns, cancellations, shipping status, and refund inquiries for Walmart.com serving millions of orders at scale.
Achieved 1-3 sec Time-to-First-Token (TTFT) for 90% calls caching FAQ responses and delivering an estimated 200,000+ hours in annual operational savings.
Fine-tuned a Policy Recommender agent using Llama-2 13B with LoRA (PEFT) for the multimodal RAG
reduced observability costs by 70% through intelligent tail-based sampling
Developed and deployed a scalable Translation App for Walmart store associates with a daily 20,000+ active users for real-time STT, TTS and TTT translations utilizing Kafka and WebSockets achieving translation and transcription in 300ms
Oct 20213 yrs 2 mosNov 2024
Bengaluru, Karnataka, India
Content Ingestion Pipeline
• Developed a RAG pipeline using Sentence Transformers for embedding, using Kubernetes Persistent Volume for reliable model loading and reducing the latency during inference
• Implemented PgVector DB and optimized Kafka, Celery, RabbitMQ, and PostgreSQL to manage and preprocess data efficiently for Dell Chat application.
• Enhanced content generation with open-source LLMs like Llama2-13B, ensuring high availability and scalability.
Cross Sell Recommendations
• Developed a machine learning model for personalized cross-sell recommendations for 30 million users, suggesting relevant SKUs based on cart items using Siamese Networks and Transformers, enhancing personalization.
• Engineered data pipelines using PySpark to handle data from Hadoop and Greenplum, with Airflow scheduling for real-time inference and task coordination.
• Utilized Redis for real-time prediction retrieval, responding to updates in users' shopping carts to deliver immediate, relevant suggestions.
• Deployed a FastAPI service on Kubernetes, integrating Redis and MongoDB to ensure scalable and reliable prediction delivery.
Predictive Modelling
Implemented LSTM deep learning models to enhance UK laptop sales by 16% through targeted brand recommendations.
Applied Double ML for accurate baseline product pricing, achieving optimal price elasticity.
Developed an XGBoost model to discern user intent on the site, improving engagement and experience.
Mar 20217 mosSep 2021
Bangalore Urban, Karnataka, India
• Employing OCR based text extraction from pdfs using Camelot. Built data pipeline to web scraped the invoice tables from client’s website and automate the process by developing a web application using Django.
• Built ETL Pipeline to efficiently process 1Billion rows of data using Pyspark and Airflow as scheduler to run the spark jobs efficiently
• Build a customizable analytics dashboard using Plotly and integrate into backend web application using Django
Feb 20165 yrs 2 mosMar 2021
Bangalore Urban, Karnataka, India
• Built classification models to predict the room, hotel type, discounts, rewards users going to buy to enhance the Guest Experiences in Hospitality Industry
• Enhanced the accuracy of the model by 10 – 15% employing different feature engineering techniques
• Decreased the latency of I/O bound operations reading from S3 Bucket using Multithreading and Caching (2 to 0.0005 sec)
• Developed topic modelling algorithm using LDA for tagging free-text in-house IT tickets to right categories and route to right agent using Spark MLlib for 900 million tickets and improved SLA from 3 days to 1 day
• Developed neural network and tree-based models to predict the arrival of a flight in +/- 5 min
• Enhanced model performance from 65 -75% by capturing the seasonality factors with respect to different time slots and for more granularity and in-depth analysis
• Built data pipleline to web scraped data from FlightStats website using Beautiful Soup to test the performance of model on real time data and model pipeline using AWS Sagemaker and monitor the metrics using CloudWatch
• Used AWS Glue and Athena for better data transformations
Midhilesh’s AI Stack
Education
Bachelor of Technology (B.Tech.), Electronics and Communications Engineering
Jawaharlal Nehru Technological University, Anantapur
20114 yrs2015
Intermediate, Science
Sri Chaitanya Junior College, Nellore
20092 yrs2011
Join Midhilesh and 10k+ other professionals on
Tal is the app where bosses meet engineers & builders directly, without middlemen
Download Tal Boss