Software Engineer
Ashu Kumawat
📍San Francisco, Bay Area
" A continuous learner and problem solver, interested in building systems that are reliable, practical, and genuinely useful."
Recent achievements:
1. 3x Hackathon Winner
2. Devpost 5+ project submission
3. Hacker rank sql gold
4. Founders School: Ideation bootcamp course completion with 100% scholorship
Year 2026 goals:
1. Participate in 20+ hackathon
2. Share Learnings on Blogs
3. Solve Daily 10+ problem on Leetcode
5. Github 500hrs+ repo building
6. Publish 2 research
Projects
As shown in example scenario Left. As soon as alert came to system. It invoke agentic flow to find safe evacuation plan under few seconds.
Disaster Alert
Fast API
Impact area mapping
Divide area into impact zones
Safe spots locator
Resource locator
Path finder
Details to User
Swarms Agentic AI flow
Agents:
-
Impact area mapping: Map total area that going to affect in this scenario.
-
Zone divider: Divide zones based on people estimations, difficulty in evacuation, and level of impact.
-
Safe Spot locator: Finds safe location to manage evacuated crowd.
-
Path finder: Finds optimal path from zone to Safe spot based on traffic, type of road and transportation available.
-
Resource allocator: Find number of current public transportations and their capacity, and give approximate buses and medical van needed.
Realtime Disaster Management
Education
California State University, East Bay
Aug 2023 - May 2025
Hayward, CA
Master in Statistics concentration Data Science
Coursework: Applied Deep learning, Machine Learning, Experimentation design, Applied statistics
Vidyalankar Institute of Technology
Bachelor in Computer Engineering
Jul 2015 - May 2018
Mumbai, India
Work Profile
Omatochi
AI Engineer
Sept 2025 - current
Hayward, CA
Myntra
Dec 2019 -Apr 2023
Machine Learning Engineer
Mumbai, India
-
Delivered caregiver and elderly support via secure health data pipelines, RAG clinical assistance, and predictive risk signals in low-latency services for in-app guidance, alerts, and care workflows.\
-
Built Python and FastAPI APIs to ingest caregiver notes, vitals, and engagement events into Postgres for care summaries, reminders, and triage.
-
Implemented an LLM RAG system using LangChain plus a vector database (ChromaDB) over approved care plans and education content to generate responses with citations.
-
Created agentic AI flow using Temporal and Redis to route caregiver messages, retrieve context, draft clinician-ready replies, and trigger follow-ups.
-
Worked on the home feed ranking model using Python, Spark, and TensorFlow to combine user signals and item features. This helped the team lift overall CTR by about 6 % on the main app feed.
-
Built the Similar Items and Complete the Look recommenders by mixing co-view data with CNN-based image embeddings in PyTorch. This improved the PDP to Add to Cart rate by around 3 % to 5 %.
-
Set up data pipelines in Spark ; Hive ; and Airflow to refresh user profiles and item embeddings every day. This cut down feature delays and kept the recs feeling more fresh for users.
-
Ran A/B tests with the product team for new ranking ideas and scoring logic. Used SQL and dashboards in Looker to read CTR, CVR, and revenue changes and decide if a model should go live.
-
Added model checks for drift and automated retrain jobs using Airflow, Python, and internal APIs. This reduced unexpected drops in rec quality and made the system more stable over time.
Happiest minds
Jun 2018 - Oct 2019
Machine Learning Engineer
Mumbai, India
-
Worked on client projects to build text classification models using Python ; scikit learn ; and simple LSTM networks in Keras. Helped the team improve accuracy on support ticket tagging by about 8%.
-
Built data cleaning and feature pipelines in Python ; Pandas ; and SQL to prepare large customer logs for ML tasks. This reduced manual cleanup time for the team and made model runs more stable.
-
Created sentiment analysis PoCs using NLTK and early transformer embeddings like ELMo for retail clients. Shared results with the consulting team to help shape the final solution.
-
Helped deploy small ML services using Flask and Docker for demo environments. Worked with DevOps to push image updates and check if the models were running fine under load.
-
Worked closely with business analysts to read client requirements and convert them into ML tasks. Built simple dashboards in Tableau to show model outputs ; trends ; and error cases to stakeholders.