Reach out for consulting engagements, production AI system design, corporate training, or product partnerships. Send the workflow, constraint, or system you want to discuss.
Discuss the system →
Senior ML Engineer and Founding AI Engineer at Vavi Labs with 6+ years in ML/AI, shipping production ML systems and building AI-native products.
Led end-to-end ownership across the MLOps lifecycle: evaluation frameworks, LLM-as-judge workflows, CI/CD regression tests, production debugging, stakeholder management, latency SLOs, drift monitoring, reliability analysis, and A/B testing. I've shipped systems across multiple domains—supervised learning, time series forecasting, anomaly detection, NLP/GenAI (RAG, LLM fine-tuning).
Building AI-native products, dev tools, plugins for coding agents, fine-tuned models, and technical education systems for engineers and AI teams.
Current Work
Systems Built
Arbiter - AI governance and decision-intelligence platform for engineering teams with CLI-first workflows, audit trails, confidence scoring, connector-backed context, and async review loops.
Creative Collab OS - beta creative AI studio for stand-up, songs, comics, sitcoms, and memes using workflow graphs, direction checkpoints, taste-memory profiles, streaming outputs, and multimodal artifact generation.
BatSwing - PWA + computer-vision cricket coaching product for one-phone side-on batting capture, pose/phase analysis, WhatsApp-ready academy reports, and consent-aware player media handling.
Tech Abstractions - AI/ML interview-prep platform focused on production reasoning, system design, LLM infrastructure, and applied practice.
Dev Tools and Illustrated Explainers - plugins for coding agents, repeatable agentic AI engineering workflows, production MLOps workflows, and illustrated AI infrastructure guides.
Fine-tuning Projects - Sitcom Scriptwriter and Kannada Physics Tutor using custom datasets, staged fine-tuning, evaluator workflows, and RAG/evaluator components.
Tech Stack: LLMs, Agentic AI, MLOps, Fine-tuning, Evaluation, Coding Agents, Product Engineering.
Built 4 production ML systems for a European e-commerce marketplace with 75K SKUs, 50K DAU, and 2.5K orders/day. Led the end-to-end ML lifecycle from problem formulation to deployment and monitoring.
Systems Built
Tech Stack: Python, LightGBM, XGBoost, AWS SageMaker, Glue, Kinesis, Lambda, Terraform, Feast, Redis, OpenAI, Mistral.
Founded two social impact ventures: an artisan-first D2C marketplace and a digital-first rural education initiative.
Spiticart - direct-to-consumer e-commerce platform supporting rural artisans across products such as Shantiniketan leather, Channapatna toys, and Agra marble carvings.
Rumi Schools - rural education initiative with curriculum across academics, technology, arts, sports, life skills, and remote mentorship as part of a larger model-village vision.
Owned end-to-end IoT ML systems for smart building operations across 3,000 apartments, spanning data quality checks, evaluation, monitoring, and scheduled retraining.
Predictive Maintenance: deployed anomaly detection and alert triage for heating systems; diagnosed alert fatigue; redesigned the workflow as a prioritization tool; evolved from residuals + LOF to supervised models as technician labels grew.
Energy Forecasting: 24-hour ahead demand forecasting with XGBoost, weather, lag, rolling-window, and holiday features; tiered cold-start strategy from physics-informed heuristics to individualized forecasts.
Tech Stack: Python, scikit-learn, XGBoost, ARIMA, PostgreSQL, Pandas, Docker.
CGPA: 5.25 / 6.0. Master's Thesis: Data Analysis and Anomaly Detection in Buildings Using Sensor Data. Focus: Machine Learning, Applied Data Analysis, Computer Vision, Natural Language Processing, and Algorithms.
Master's Thesis: Data Analysis and Anomaly Detection in Buildings Using Sensor Data.
Conducted computer vision research for real-time object detection and scene understanding in traffic surveillance systems.
Traffic Surveillance Object Detection: fine-tuned YOLOv2, tuned anchor boxes/heads, and used targeted augmentation to surface candidate "tandem near pedestrian" incident windows for human review.
Scene Understanding: built a PyTorch subject-predicate-object detector with scene-graph outputs; engineered spatial features and explored graph/message-passing plus translation-embedding formulations for more robust predicate classification.
Tech Stack: Python, PyTorch, YOLOv2.
Developed advanced signal processing algorithms for WCDMA/LTE wireless base stations. Joined as the fifth employee at the semiconductor startup. Developed and validated WCDMA uplink receiver algorithms: Path Searcher, RAKE, and Multiuser Detection.
US Patent 9602240: symbol-level interference cancellation at a receiver for multiuser detection.
Key Achievements
Patents
US-9602240-B1: symbol-level interference cancellation for WCDMA uplink receivers using soft-bit estimates across iterative decoding.
US-20160365991-A1: two-stage channel estimation that reduces correlation operations by limiting fine search to regions of interest.
Tech Stack: MATLAB, Octave, Signal Processing, WCDMA, LTE.
Conducted research on diffuse optical tomography at IISc Bangalore. Developed a data-resolution matrix method from the sensitivity/Jacobian model and regularization to identify independent measurements and published a first-author paper in Medical Physics.
Research Contributions
Tech Stack: MATLAB, optimization algorithms, inverse problems.
CGPA: 9.23 / 10.0. Focus: Signal Processing, Communications, Algorithms, and Mathematics.
Built a strong foundation in mathematics, signal processing, communications, and algorithms that provided the technical base for later work in ML, wireless communications, and systems engineering.
