Ankit RajAI Engineer — Agentic Systems, RAG & Applied ML

A production ReAct agent with a hardened code sandbox, on-demand skills and speech-to-speech voice.

• LangGraph
• FastAPI
• Qdrant
• Redis
• MongoDB
• Docker
• AWS
• gVisor

• Production agentic workspace built on a single ReAct LangGraph agent — code sandbox, on-demand skills library, agentic RAG, MCP tool integration, persistent memory and speech-to-speech voice (~31K lines of Python, 185 commits, 167 tests).
• Decoupled agent execution from the HTTP request via an in-process turn registry that fans one running turn to N subscribers and replays history to late attachers — closing a tab no longer kills generation, and a reload reattaches mid-turn.
• Hardened the code sandbox after self-directed red-teaming, moving from in-process monkeypatching to a dedicated non-root UID with --cap-drop=ALL plus an optional gVisor-isolated remote execution host that fails closed.

End-to-end MLOps pipeline with automated CI/CD, a cloud model registry and a deployed inference API.

• MLOps
• Docker
• AWS
• FastAPI
• MongoDB

• Built end-to-end MLOps pipeline predicting customer interest in vehicle insurance using 12 features.
• Implemented automated CI/CD with GitHub Actions, Docker, AWS ECR, and EC2 for model deployment.
• Deployed FastAPI service with S3 model registry, MongoDB Atlas, and schema-driven data validation.

A decoder-only language model built from scratch: modern architecture, pretraining, instruction tuning and preference alignment.

• PyTorch
• LLM
• RoPE
• GQA
• Deep Learning

• Built TinyLLM with WordPiece, RoPE, GQA attention, GELU, and pre-LayerNorm in a decoder-only GPT model.
• Trained with mixed-precision next-token LM using AdamW + warmup for efficient single-GPU optimization.
• Added instruction tuning and preference ranking alignment instead of RLHF to refine model behavior.

A risk-calibrated default classifier tuned to the cost asymmetry of missing a defaulter.

• XGBoost
• SMOTE
• Risk Modeling
• Finance

• Built a predictive score using XGBoost on 30,000+ records, addressing imbalance via SMOTE.
• Achieved F2-score of 0.603, identifying 84% of defaulters for early warning and loss mitigation.
• Created 8 financial features and tuned threshold aligned with the bank's risk policy.

Deep learning models forecasting firm-level market cap across multi-year horizons.

• LSTM
• Encoder–Decoder
• PCA
• Time Series

• Built forecasting models to predict firm-level market capitalization across 1–3 year horizons.
• Developed LSTM, MLP, and Encoder–Decoder architectures on 25 years of historical market data.
• Applied PCA on 28 indicators to reduce dimensionality and improve model generalization.

ConvLSTM over satellite time series, forecasting land-cover change across two Indian cities.

• ConvLSTM
• Remote Sensing
• Computer Vision

• Employed ConvLSTM model on satellite data to predict urban expansion, monitoring land cover changes over time.
• Collected Remote Sensing Time Series satellite data for 2 Indian smart cities, Mumbai and Pune, for a 15 year time period.
• Evaluated the model using a confusion matrix, attaining an F1 score of 0.81 and 72% prediction accuracy.

Delay severity forecasting over 180K flights, with SHAP-driven operational recommendations.

• LightGBM
• XGBoost
• SHAP

• Performed EDA on 180K flight records, built predictive models using LightGBM and XGBoost with custom OAI logic.
• Achieved 0.73 ROC-AUC, 0.61 F1-score, and ~30 min MAE for delay severity forecasting.
• Used SHAP for interpretability and proposed strategies to reduce controllable delays across operations.