Lalit Shirsath — Data Scientist

Email: lalitshirsath1111@gmail.com Myself Lalit Shirsath, Data Scientist at Infocusp Innovations. I completed computer engineering I feel my sincere and dedicated approach coupled with my hunger for learning and adapting to any new technology will enable me to achieve the required outcomes and deliver success for any organisation that I work for. Currently at Infocusp Innovations, I am exploring the data science field. • Worked on LLMs like GPT3.5, GPT4, Llamma3 to perform Text Classification and automate Data Annotation Tasks with 95% accuracy and successfully use them in production to reduce the human efforts with different Prompt Engineering Techniques like Zero-Shot, Few-Shot, Chain of Thought, Chain of Logic, ReAct prompt • Built a legal-domain chatbot using a RAG architecture to answer queries grounded in uploaded contracts • Handled semi-structured legal documents (tables, lists, clauses) by preserving structure during parsing using Unstructured.io so extraction does not break tables or hierarchical context • Built a hybrid retrieval system that fuses BM25 and embedding scores, then applies re-ranking to surface the most relevant candidates. Employed a MultiVectorRetrieval to store and search clause text, summary, table view embeddings for better coverage and recall • Integrated with a scalable vector database using hybrid / semantic search capabilities, combining with HNSW indexing for low-latency retrieval. Optimized system on Query Expansion, Prompt Compression and Metadata Filtering to boost retrieval precision. Integrated RAGAS for evaluation. Used Precision@k, Recall@k, MRR, Faithfulness, Hit Rate, Context and Answer Relevancy and Correctness for ensuring high-quality and trustworthy responses. • Built and optimized NLP/ML pipelines for legal contracts, achieving 93%+ F1-score, improving pipeline speed by 30%, developing 50+ SVM classifiers, and deploying scalable workflows using Temporal, FastAPI, Docker, LiteLLM. Technical Skills: Programming & Tools: :Python, SQL, Data structures, Algorithms, Problem Solving, Nvidia CUDA, FastAPI. ML Technologies: Machine Learning, Deep Learning, Generative AI, NLP LLM Techniques/Frameworks: Langchain, Llamaindex, RAG Cloud Technologies: Cloudformation, EC2, Lambda, ECS, ECR, S3, API Gateway, Step Functions. Achievements • Solved 1000+ problems on across coding platforms • Secured 2nd Rank in Code-O-Fiesta 2021 coding competition from 300+ participants. • Institute Rank 5 in GeeksForGeeks from 3000+ students • Secured 9th Rank in THE ENIGMA WAR coding competition 2022. Resume attached below

Stackforce AI infers this person is a Data Scientist specializing in NLP and Machine Learning within the LegalTech and SaaS industries.

Location: Pune, Maharashtra, India

Experience: 3 yrs 4 mos