Jan 2025 – Present · 1 yr 4 mos

Jul 2023 – Present · 2 yrs 10 mos

• Engineered an LLM-powered (OpenAI and OpenSouce Models), Queue and
• API-based text generation pipeline via Python from scratch.
• Implemented Factory, Strategy, Decorator, Template Design Pattern, etc by
• adhering SOLID principles to create a modular codebase.
• Leveraged Redis, Aerospike, and MySQL databases, along with Kafka, to build
• and optimize scalable data pipelines in languages like Python and Java.
• Executed Supervised Fine-tuning on a large-scale dataset of 1.5 million rows,
• Utilizing Parameter Efficient Fine-tuning Techniques, LoRA to optimize training.
• Explored PPO and DPO Algorithms of Reinforcement Learning with Human
• Feedback (RLHF) for incorporating preferential behavior LLM.
• Deployed Fine-tuned models and exploited consistent batching through
• LoRAX to improve the throughput by 4 times.

Jan 2023 – Jun 2023 · 5 mos

• Developed an NLP-based Granular keyword classification algorithm.
• Utilised Elastic search and Wikipedia database to search keywords in real-time.
• Extracted category taxonomy (5 times bigger than existing taxonomy) from
• Wikipedia using Beautiful soup and Graph Algorithms.
• Implemented a Ranking algorithm (combination of Frequency based and
• Hierarchical Ranking) to rank retrieved categories.
• Utilized Multi-Threading to reduce the run-time of the algorithm by 80%

Jan 2023 – Jun 2023 · 5 mos