Pratham Uday G profile photo 1 on Tal

Pratham Uday G

Generative AI , Machine Learning, Data Science

Bengaluru

1+ year of work experience

TURYON Tech Labs Pvt. Ltd
Aug 202510 mosMay 2026

Designed and implemented a hybrid code-understanding system SyntaxLens combining rule-based AST parsing with LLM-driven semantic inference to explain Python code at three configurable depth levels prioritizing structural precision and reduced hallucination compared to pure LLM approaches. Built a RAG retrieval layer that injects language-specific context (Java idioms, Python-to-Java mapping rules) at inference time using FAISS-backed vector search, improving explanation accuracy without retraining the base model. Designed a multi-step prompt engineering pipeline with structured output validation; enforcing schema checks on LLM responses before surfacing to users reduced incorrect code mappings by ~25–30% (based on evaluation suite metrics). Built an automated evaluation harness scoring explanation correctness and mapping fidelity using rule-based validation and semantic similarity checks across a curated test suite, enabling measurable, iterative prompt refinement with clear improvement signals. Optimized chunking strategy and context window management to achieve sub-3s end-to-end response latency for Python functions up to 150 lines navigating the trade-off between retrieval depth and token cost to stay within API budget constraints. Exposed the system via a lightweight REST API layer, enabling integration with a frontend interface for real-time code explanation queries by end users. Engineered failure-handling logic for edge cases (ambiguous mappings, out-of-context snippets), surfacing graceful fallback explanations instead of raw LLM errors improving perceived system reliability.
Intellipaat
Feb 20251 yr 2 mosMar 2026

Bengaluru, Karnataka, India

Applied advanced machine learning and Deep Learning algorithms (Linear Regression, Logistic Regression, Decision Tree, Random Forest, Bagging, Agglomerative Clustering, AdaBoost, Gradient Boost, XGBoost) , ANN, CNN, RNN , LSTM ,GRU on Kaggle datasets, performed in-depth EDA, improved model accuracy, F1, Recall Confusion Matrix and created insightful visualizations for data-driven decision-making.
Intellipaat
Feb 20256 mosJul 2025

Executed multiple end-to-end ML projects as part of a structured, mentor-guided industry training program. Built and iterated end-to-end ML pipelines across 4+ projects spanning data ingestion, preprocessing, feature engineering, model training, and evaluation against production-style KPIs. Engineered features from structured and noisy real-world datasets; applied cross-validation and hyperparameter search (GridSearchCV, RandomizedSearch) to reduce overfitting and improve model generalization. Integrated trained models into MS SQL Server reporting workflows, enabling near-real-time predictions accessible to business stakeholders without ML backgrounds. Version-controlled experiments with Git and maintained reproducible notebooks; documented model iterations and evaluation outcomes for team handoff and review cycles.

Education

Presidency University Bangalore

Bachelor of Technology - BTech, Information Science and Engineering

Presidency University Bangalore

20223 yrs2025

Join Pratham and 10k+ other professionals on

Tal Boss

Tal is the app where bosses meet engineers & builders directly, without middlemen

Download Tal Boss

Pratham Uday G on Tal

Headline

Generative AI , Machine Learning, Data Science

About

Information Science and Engineering graduate with a strong foundation in Data Science, Artificial Intelligence, and Machine Learning. Passionate about leveraging data, technology, and problem-solving skills to deliver innovative solutions and drive meaningful impact in forward-thinking organizations. Eager to contribute to projects that harness data-driven insights, intelligent automation, and cutting-edge AI/ML techniques to solve real-world challenges.

Location

Bengaluru

Work experience

Machine Learning Engineer at TURYON Tech Labs Pvt. Ltd, Aug 2025 - May 2026 · 10 mos, Designed and implemented a hybrid code-understanding system SyntaxLens combining rule-based AST parsing with LLM-driven semantic inference to explain Python code at three configurable depth levels prioritizing structural precision and reduced hallucination compared to pure LLM approaches. Built a RAG retrieval layer that injects language-specific context (Java idioms, Python-to-Java mapping rules) at inference time using FAISS-backed vector search, improving explanation accuracy without retraining the base model. Designed a multi-step prompt engineering pipeline with structured output validation; enforcing schema checks on LLM responses before surfacing to users reduced incorrect code mappings by ~25–30% (based on evaluation suite metrics). Built an automated evaluation harness scoring explanation correctness and mapping fidelity using rule-based validation and semantic similarity checks across a curated test suite, enabling measurable, iterative prompt refinement with clear improvement signals. Optimized chunking strategy and context window management to achieve sub-3s end-to-end response latency for Python functions up to 150 lines navigating the trade-off between retrieval depth and token cost to stay within API budget constraints. Exposed the system via a lightweight REST API layer, enabling integration with a frontend interface for real-time code explanation queries by end users. Engineered failure-handling logic for edge cases (ambiguous mappings, out-of-context snippets), surfacing graceful fallback explanations instead of raw LLM errors improving perceived system reliability.

Data Scientist at Intellipaat, Feb 2025 - Mar 2026 · 1 yr 2 mos, Bengaluru, Karnataka, India, Applied advanced machine learning and Deep Learning algorithms (Linear Regression, Logistic Regression, Decision Tree, Random Forest, Bagging, Agglomerative Clustering, AdaBoost, Gradient Boost, XGBoost) , ANN, CNN, RNN , LSTM ,GRU on Kaggle datasets, performed in-depth EDA, improved model accuracy, F1, Recall Confusion Matrix and created insightful visualizations for data-driven decision-making.

Intern at Intellipaat, Feb 2025 - Jul 2025 · 6 mos, Executed multiple end-to-end ML projects as part of a structured, mentor-guided industry training program. Built and iterated end-to-end ML pipelines across 4+ projects spanning data ingestion, preprocessing, feature engineering, model training, and evaluation against production-style KPIs. Engineered features from structured and noisy real-world datasets; applied cross-validation and hyperparameter search (GridSearchCV, RandomizedSearch) to reduce overfitting and improve model generalization. Integrated trained models into MS SQL Server reporting workflows, enabling near-real-time predictions accessible to business stakeholders without ML backgrounds. Version-controlled experiments with Git and maintained reproducible notebooks; documented model iterations and evaluation outcomes for team handoff and review cycles.

Education

Bachelor of Technology - BTech, Information Science and Engineering, Presidency University Bangalore