Dec 2024 – Present
- Built multilingual agentic LLM with RAG + BiLSTM supporting farmers across 12+ regions
- Deployed real-time maize pricing & buyer-seller matchmaking via ONDC & eNAM
- Fine-tuned Florence-2 for multimodal crop quality analysis
- Achieved 92% classification accuracy using OpenCV pipelines
- Scaled on Microsoft Azure infrastructure
Sep 2024 – Dec 2024
- Built Python-based LLM pipelines on Clarifai achieving 30% efficiency gain
- Engineered Chrome extension for automated web scraping
- Extracted structured data from 10+ websites
- Reduced manual data collection by 70%
May 2023 – Sep 2023
- Enhanced image preprocessing pipelines achieving 15% better downstream performance
- Optimized deep learning models in MATLAB
- Achieved +18% accuracy and –20% training time
Tech Stack: Python StableBaselines3 PyBullet OpenAI Gym
Achievements:
- Multi-agent RL coordination with battery-aware reward shaping
- 90%+ success rate across diverse simulation scenarios
- Real-time obstacle avoidance and dynamic target tracking
Tech Stack: Next.js Python Firebase NeonDB LangChain
Features:
- Adaptive interview system using real-time AI feedback
- Intelligent resume parsing + job matching algorithms
- 50% increase in session completion rates
- Avg 15 relevant jobs surfaced in first week
Tech Stack: Next.js Redis Python Vercel TensorFlow
Impact:
- RL-based AI therapist trained on 10,000+ conversations
- 85% positive user feedback across all sessions
- Features: Journaling, breathing exercises, focus tools, community
- 40% engagement growth in first month
AI Model: Custom PPO-trained conversational agent with empathy scoring
