Projects
Projects
Developed an Agentic RAG system using LangChain & LangGraph to retrieve, analyze, and evaluate research papers via LLM agents on GCP.
Feb 2025 - May 2025 / Github
Built a Kafka–PySpark pipeline for financial news analysis using LLaMa and RAG, improving accuracy and reducing latency by 40%.
Sep 2024 - Dec 2024 / Github
Fine-tuned LLaMa-2 with DPO/ORPO for safety on PKU-SafeRLHF and achieved 93% safety using Llama-Guard and LLM-as-a-judge.
Feb 2024 - May 2024 / Github
Generated text-to-motion via GANs and VAE on HumanML3D, achieving 91% FLOPs reduction and FID of 0.48 over latent diffusion models.
Feb 2024 - May 2024 / Github
Implemented and benchmarked advanced RL algorithms like PPO, Actor-Critic, and n-step SARSA across classic control tasks in PyTorch.
Sep 2023 - Dec 2023 / Github
Developed transformer-based gyro correction models surpassing CNN baselines in orientation tracking on EUROC dataset.
Apr 2021 - Jul 2021 / Github
Used Grad-CAM, Fourier analysis, and complex-valued nets to explain adversarial behavior in CNNs across image domains.
Jan 2021 ‑ Apr 2021 / Github
Built a vision-language model with attention-based decoding and GANs to caption and predict missing video frames using Optical Flow.
Sep 2020 ‑ Dec 2020 / Github
Designed a competitive RL agent for Connect-4 using MCTS and Actor-Critic, with transfer learning for higher-dimensional boards.
Jan 2020 ‑ Apr 2020 / Github
Built and deployed few-shot facial recognition using Siamese, Prototypical, and Relation networks with OpenVINO optimization.
Jan 2020 ‑ Apr 2020 / Github
Designed a gesture-based digital writing device using accelerometer data and trained with SVMs and ANN for character recognition.
Jan 2019 ‑ Apr 2019 / Github