|
Innovative Technology Implementation of the Year
|
Excellence in Data Innovation (Internet Services)
|
|
|
role: "Data Scientist & ML Engineer"
specialization: "Production-scale LLM Systems & GenAI Infrastructure"
current: "Building LLM Infrastructure @ Shaadi.com | Astrochat.com"
impact:
- "3M+ users served"
- "100K+ daily LLM API calls"
- "99.9% uptime maintained"
- "40% cost reduction achieved"
- "Multiple Production GenAI features shipped"
expertise:
- "LLM Fine-tuning (LoRA/PEFT)"
- "Inference Optimization (vLLM)"
- "Multi-agent Orchestration (LangGraph)"
- "Production MLOps"
Built enterprise LLM gateway handling 2K+ concurrent requests with intelligent routing, caching, and cost monitoring
|
Fine-tuned & distilled LLMs achieving 40% better relevance while reducing model size by 65%
|
Deployed 7 GenAI features with measurable business impact: 2× F2P conversion, 25% engagement boost
|
Specialized: vLLM • LangGraph • LiteLLM • LoRA/PEFT • RAG • Model Distillation
Tools: Langfuse • MLFlow • Datadog • CloudWatch • CI/CD
🔥 Stock Price Prediction - Transformers ⭐ 50+ Stars
Built time-embedding attention model for price forecasting, achieving 18% accuracy improvement. Open-sourced for community use.
Tech: Python • Transformers • PyTorch • Time-Series Analysis
Production-grade LLM ecosystem powering 3M+ users:
- LLM Gateway: 2K+ concurrent requests, intelligent routing, Redis caching (35% hit rate)
- Observability: Self-hosted Langfuse for prompt versioning, evaluation, monitoring
- Fine-tuning Pipeline: LoRA/PEFT on 3M+ user profiles, 40% relevance improvement
- Multi-agent Orchestration: LangGraph-based engine managing 200K+ monthly conversations
Impact: 99.9% uptime • 40% latency reduction • 20% cost savings • 7 GenAI features shipped
Formulated Deep Reinforcement Learning approach for image restoration using POMDP framework. Achieved 12% PSNR improvement over traditional methods.
Tech: Python • PyTorch • Deep RL • Computer Vision
Multi-person pose estimation system using TensorFlow MobileNet on COCO dataset. 20+ FPS video inference with 92% keypoint detection accuracy.
Tech: TensorFlow • MobileNet • OpenCV • Real-time Processing
Indian Institute of Technology, Jodhpur (2020 - 2024)
B.Tech in Artificial Intelligence and Data Science
Relevant Coursework: Machine Learning • Deep Learning • Computer Vision • Advanced ML • Optimization for ML • Data Engineering
Data Scientist | Jul 2024 - Present | Mumbai, India
Key Achievements:
- 🏗️ Architected production LLM ecosystem serving 3M+ users with 100K+ daily API calls
- 🚀 Deployed custom LLMs achieving 40% latency reduction and 20% cost savings
- 🎯 Improved F2P conversion by 2× (7% → 14.2%) through data-driven prompt optimization
- 👥 Mentored 4 junior data scientists on LLM fine-tuning and production patterns
- 🔧 Standardized company-wide LLM access by migrating 5 services to centralized infrastructure
Data Science Intern | Jun 2023 - Aug 2023 | New Delhi, India
- 📊 Processed 1.2M QnA dataset using Databricks and PySpark with 99.9% uptime
- 🎯 Built NLP quality model boosting answer accuracy by 19.81% (73.5% → 93.3%)
- 🔍 Reduced duplicate questions by 35% using Siamese Neural Network
- 👔 Vice President - Board of Students' Welfare, IIT Jodhpur (2023)
- 🏆 8th Position - Inter IIT Tech Meet (National-level competition)
- ♟️ Interests: Chess • Music • Gaming • Linux Enthusiast
Open to opportunities in:
ML Engineering • LLM Infrastructure • AI Research • Staff/Principal Engineer Roles
