Surya.
I build, break, and learn.
AI/ML Engineer building production-grade Machine Learning and AI systems, with expertise in LLMs, Agentic AI, RAG, multimodal AI, and scalable backend infrastructure.
Hi, My name is
About Me
I’m a Physics postgraduate who accidentally wandered into AI and decided it was too interesting to leave.
I enjoy building things that solve real problems, whether that’s training machine learning models, orchestrating LLMs, or convincing autonomous AI agents to behave themselves (with mixed success). Over the years, I’ve worked across traditional machine learning, Generative AI, RAG, and Agentic AI, turning ideas into systems that people can actually use.
What keeps me hooked isn’t a particular technology; it’s the process of figuring things out. I like breaking complex problems into simpler ones, learning whatever I need along the way, and building something better than what I started with. If there’s a rabbit hole worth going down, there’s a good chance I’ve already opened fifteen tabs about it.
Outside of work, you’ll probably find me planning my next trip, catching up on anime, or hunting for places that serve great food.
Here's a glimpse of my current tech stack:- Python
- SQL
- Machine Learning
- Deep Learning
- Computer Vision
- Agentic AI
- LLMs
- RAG
- AWS
- FastAPI
Experience
- Led the development of an AI-powered spiritual guidance platform, combining semantic routing, LLM orchestration, and AI safety mechanisms to deliver reliable, personalized conversations while mitigating prompt injection and adversarial inputs.
- Implimented a production-grade RAG architecture using Amazon Bedrock foundation models, HuggingFace embeddings, and a domain-specific Bhagavad Gita knowledge base to generate grounded, context-aware responses with high retrieval accuracy.
- Created a multimodal AI video intelligence platform integrating predictive ML models, Whisper ASR, and Azure Claude to analyze audience retention, identify engagement bottlenecks, and generate actionable editorial reports and timestamped script revisions.
- Co-founded and architected an AI-driven AIOps platform for observability, anomaly detection, correlation-based RCA, and agentic AI analytics across enterprise infrastructure.
- Built EnviState and EnviDetect, delivering real-time health scoring and hybrid Weak–Strong learning for anomaly detection and forecasting, predicting incidents 30–60 minutes ahead of SLA impact while reducing compute costs.
- Developed EnviRelate, a correlation engine, leveraging service dependency graphs to identify root causes and cascading failures, reducing investigation effort by ~40%, alongside an LLM-powered analytics platform with semantic routing, Text-to-SQL, and AI safety guardrails.
- Architected scalable ML pipelines, agentic workflows, and streaming systems, leading end-to-end development from design to production.
- Designed and implemented a precision agriculture analytics platform using multi-year Sentinel-2 imagery, MSAVI2, and 5-year Maximum Value Composite (MVC) processing to identify high-productivity management zones while minimizing weather effects.
- Developed geospatial ML pipelines using Fuzzy C-Means clustering, hemisphere-aware seasonal logic, and Fuzzy Partition Coefficient optimization to generate adaptive management zones.
- Delivered an AI-driven deforestation compliance system using agent-based workflows to evaluate farmer documentation, identify compliance gaps, and generate remediation guidance for European export regulations, significantly reducing manual review effort.
- Worked on building and optimizing the state-of-the-art Machine learning and Deep learning models.
- Learned how to integrate Deep Learning techniques, such as CNN & OpenCV, with Astrophysics.
- Worked on Galaxy Catagorization.
- Built a Model to Detect Anomalies in Surveillance videos using Neural network architectures.
Education
2020 - 2022
M.Sc. Physics
Utkal University, Bhubaneswar, Odisha
GPA:
7.6 out of 10.0
During my dissertation project, I worked upon Detecting Exo-Planets using Machine Learning.
- Built SVM classifier for Time series data.
- Used Pre-processing technique “FFT” on Time series to improve the model performance.
Projects
OpenCV
Tensorflow
Keras
Deep Neural Network
Anomaly Detection in Surveillance videos
A Neural network architecture to detect anomalies like accident or burglary in surveillance videos.
ML
EDA
Streamlit
Model Building
Multiple Disease Prediction
A web app to predict whether someone has health issues or not. Currently this app only predicts 3 types of diseases i.e.- "Diabetes", "Heart disease" & "Parkinson's disease".
Machine Learning
EDA
SVM
Feature Engineering
Model Building
Exo-planet Detection using Machine Learning
The aim of this Project was to Detect Exo-planets from light curves of Kepler Mission using FFT and analyze to what extent they affect the accuracy of Exo-planets Classification.
Achievements
HackerRank Gold Badge in Python
Received Gold badge from HackerRank by solving various problems in python.
INSPIRE Scholarship Holder
Funded by "DST, Government of India" to the top 1% scorer in 12th examination in state division.
IIT-JAM, NEST Qualified
Qualified some national level entrance tests in physics.
Get in Touch
My inbox is always open. You can ping me here!