5+ yrs
building and shipping production ML, Generative AI, and data-science systems across financial services
New Jersey · Open to senior AI/ML roles
Shipping ML systems at the
edge of finance, risk,
and retrieval.
I'm Sai Nikhil — an AI/ML engineer and data scientist with 5+ years building production ML and Generative AI systems. Currently a Senior AI Engineer at Databricks, building enterprise Generative AI and Knowledge Assistant capabilities — RAG pipelines, agentic workflows, and governed, production-grade LLM serving. Previously at Goldman Sachs, building financial research and risk-intelligence platforms.
35%
higher information-retrieval accuracy in production RAG pipelines for financial research
25%
improvement in risk-prediction and fraud-detection model accuracy via tuning + feature work
40%
faster query response across a 100 TB+ analytics warehouse for risk and regulatory reporting
— 00
Hello — I'm Sai Nikhil.
I build production ML and Generative AI systems. My day-to-day at Databricks is shipping enterprise Generative AI and Knowledge Assistant capabilities — RAG pipelines over Databricks AI Search, agentic workflows with Unity Catalog Functions, and governed real-time model serving.
Before Databricks, I spent two and a half years at Goldman Sachs as an AI/ML Engineer building financial research and risk-intelligence platforms — RAG pipelines, agentic LLM workflows, and models for risk prediction, anomaly detection, and fraud. Before that, two and a half years at Cognizant as a Data Scientist — building real-time market models, reinforcement-learning execution strategies, anomaly-detection frameworks, and NLP sentiment pipelines over large-scale financial data. I hold a Master's in Computer Science from SUNY Albany and a Bachelor's in Computer Science from Bharath University, India.
My work sits where modeling meets engineering: latency budgets, evaluation that survives production, and pipelines that don't fall over when the data changes underneath them. Offline accuracy is the start of the job — shipping is the rest.
- Role
- Senior AI Engineer · Databricks
- Based in
- New Jersey, USA
- Education
- M.S. CS · SUNY Albany
- Specialties
- Generative AI · RAG · Financial ML · MLOps
- Open to
- Senior AI/ML Engineer roles
- Contact
- sai15121104@gmail.com
Work history
Five-plus years across financial services and enterprise data science — from EDA to production GenAI.
Databricks
Senior AI Engineer
Building enterprise Generative AI and Knowledge Assistant capabilities on Databricks and AWS — governed RAG pipelines, agentic workflows, and real-time model serving over structured and unstructured enterprise data.
- Built enterprise Generative AI and Knowledge Assistant capabilities on Databricks and AWS, enabling real-time conversational access to governed enterprise documents and structured data.
- Developed Python, PySpark, SQL, and Lakeflow pipelines on AWS S3 to ingest, clean, chunk, and enrich enterprise documents — 18% lower data-processing time.
- Curated AI-ready datasets in Delta Lake and managed access through Unity Catalog, implementing RBAC, lineage, auditing, and PII controls across governed data and AI assets.
- Built RAG pipelines with Databricks AI Search and embeddings — semantic retrieval, metadata filtering, and reranking for 16% higher retrieval relevance.
- Integrated LLMs and Foundation Models with prompt engineering, grounding, and response validation to generate context-aware enterprise responses and mitigate hallucinations.
- Developed agentic workflows with Knowledge Assistant, MCP, and Unity Catalog Functions, letting the assistant securely retrieve information and invoke approved enterprise tools.
- Deployed real-time inference through Databricks Model Serving and Databricks Apps, exposing scalable REST APIs with 99%+ service availability.
- Tuned retrieval, indexing, prompts, and model-serving configurations — 25% lower end-to-end response latency based on production monitoring.
- Databricks
- PySpark
- Delta Lake
- Unity Catalog
- RAG
- MCP
- MLflow
- AWS
Goldman Sachs
AI/ML Engineer
Building AI systems for financial research and risk intelligence — RAG-based knowledge discovery, agentic LLM workflows, and ML models for risk, anomaly, and fraud detection across large-scale financial data.
- Built an AI-powered Financial Research & Risk Intelligence Platform with Python, LangChain, Azure OpenAI, and Hugging Face Transformers to automate financial knowledge discovery and analysis.
- Designed RAG pipelines with LangChain, OpenAI embeddings, and vector databases — 35% higher information-retrieval accuracy for financial research and knowledge discovery.
- Engineered scalable ETL pipelines with PySpark, SQL, and Apache Spark to process over 2 TB of structured and unstructured financial data.
- Built ML models for risk prediction, anomaly detection, and document classification (Scikit-learn, TensorFlow) — 25% higher accuracy for risk-assessment and fraud-detection workflows.
- Accelerated model training and large-scale data processing with JAX, cutting experimentation and training time by 30%.
- Built multi-step reasoning workflows with LLM agents and tool-calling to automate financial KPI analysis and investment research; fine-tuned HF Transformers for 25% better extraction and classification.
- Deployed ML and GenAI apps on AWS with Docker, MLflow, and CI/CD, orchestrated with AWS Lambda / Step Functions and Vertex AI Pipelines — 35% faster model lifecycle.
- Python
- LangChain
- Azure OpenAI
- HuggingFace
- PySpark
- JAX
- Scikit-learn
- TensorFlow
- AWS Lambda
- MLflow
- Vertex AI
Cognizant
Data Scientist
Built real-time ML and data-science systems for trading, execution, and risk — market models, RL execution strategies, anomaly detection, and NLP sentiment over large-scale financial data.
- Developed and deployed a real-time ML model (Python, TensorFlow, scikit-learn) for market-fluctuation analysis over price trends, order-book dynamics, and macro indicators — 15% better predictive accuracy.
- Engineered an RL-driven execution model (OpenAI Gym, TensorFlow) that adapts trade placement to liquidity and market microstructure — 5% lower transaction costs.
- Designed an anomaly-detection framework with Apache Kafka and PySpark to flag market manipulation in real time — 30% fewer false positives, aligned to SEC & FINRA standards.
- Optimized SQL on a 100 TB+ Azure Synapse warehouse (partitioning, distribution, indexing) — 40% faster query response for risk and regulatory reporting.
- Led statistical experiments (NumPy, SciPy, StatsModels) on new trading signals — 10% better market-making and arbitrage efficiency.
- Built interactive Tableau dashboards for live trading metrics, risk exposure, and PnL — 25% faster stakeholder decisions.
- Designed an NLP pipeline (BERT, NLTK) over earnings reports and financial news — 18% higher sentiment-prediction accuracy for risk teams.
- Python
- TensorFlow
- scikit-learn
- OpenAI Gym
- Kafka
- PySpark
- Azure Synapse
- BERT
- NLTK
- Tableau
Signature projects
Production ML systems and open-source work where the architecture, the model, and the outcome line up.
P/01
Enterprise Knowledge Assistant & Governed RAG
Real-time conversational access to governed enterprise documents and structured data. Lakeflow + PySpark ingestion into Delta Lake, Unity Catalog governance (RBAC, lineage, PII controls), Databricks AI Search retrieval, and agentic workflows via MCP and Unity Catalog Functions — served through Databricks Model Serving.
16%retrieval relevance
25%lower latency
99%+availability
- Databricks
- Delta Lake
- Unity Catalog
- RAG
- MCP
- MLflow
P/02
Financial Research & Risk Intelligence Platform
End-to-end platform that automates financial knowledge discovery and risk analysis. Python + LangChain + Azure OpenAI + Hugging Face Transformers over a 2 TB+ corpus of structured and unstructured financial data, with Scikit-learn / TensorFlow models for risk prediction, anomaly detection, and document classification.
35%retrieval accuracy
25%risk-model accuracy
2TB+data processed
- Python
- LangChain
- Azure OpenAI
- HuggingFace
- Scikit-learn
- TensorFlow
P/03
RAG Financial Knowledge Discovery
Agentic RAG system for financial research — LangChain with OpenAI embeddings over vector databases, plus multi-step LLM reasoning and tool-calling that automate KPI analysis and investment-research workflows. Fine-tuned Transformers sharpen extraction and classification.
35%retrieval accuracy
25%extraction lift
Agentstool-calling
- LangChain
- OpenAI
- Vector DB
- Agents
- RAG
- PEFT
P/04
Real-time Anomaly Detection & RL Execution
Real-time financial systems for execution and surveillance — an RL-driven execution model (OpenAI Gym + TensorFlow) tuned to liquidity and market microstructure, and a Kafka + PySpark anomaly-detection framework flagging market manipulation in real time, aligned to SEC & FINRA standards.
30%fewer false positives
5%lower transaction cost
SECFINRA aligned
- Apache Kafka
- PySpark
- OpenAI Gym
- TensorFlow
- Anomaly Detection
Open source · selected GitHub repositories
All repos on GitHub →
-
self-tool-agent Python
Agent that routes each request to a vetted pre-built tool (weather, currency, search, translation, calendar) or, for pure computation with no existing tool, safely writes and sandbox-tests its own function for reuse — served as a FastAPI web app with CI and Docker.
-
resilient-llm-gateway Python
Resilience middleware for LLM API calls — retries with exponential backoff, automatic provider failover, and a circuit breaker that skips a provider known to be down, all config-driven and verified with fault injection and stress testing.
-
self-healing-rag Python
LangGraph RAG pipeline that critiques its own answers for groundedness, rewrites and retries the query on rejection, and refuses out-of-scope questions before retrieving — evaluated end-to-end against a real SEC 10-K filing.
-
rag-from-scratch JavaScript
Retrieval-Augmented Generation, implemented from first principles — chunking, embedding, indexing, retrieval, and generation without framework abstractions.
-
rag-chatbot Python
Production-shaped RAG chatbot — vector retrieval, prompt orchestration, and a pluggable LLM layer with the kind of structure you actually deploy behind an API.
-
langchain-rag-document-understanding Jupyter
LangChain-based RAG for document understanding — chunking strategies, retriever tuning, and grounded Q&A across heterogeneous corpora, captured in reproducible notebooks.
-
mlops-app HCL
MLOps reference stack — Terraform-managed infrastructure for training, model registry, deployment, and monitoring; the boring infrastructure that makes ML actually shippable.
Technical stack
Tools I've shipped with — grouped by the part of the system they serve.
Languages
- Python
- SQL
- R
ML & Deep Learning
- Scikit-learn
- TensorFlow / Keras
- Random Forests
- SVM
- Neural Nets
- K-Means / KNN
- JAX
Generative AI
- LangChain
- OpenAI API
- Hugging Face
- Azure OpenAI
- RAG
- Prompt Eng
- Embeddings
- LoRA / PEFT
Data & Databases
- Pandas
- NumPy
- SciPy
- PySpark
- Kafka
- FAISS / Pinecone
- PostgreSQL
- MongoDB
- Delta Lake
- Unity Catalog
Cloud & Pipelines
- AWS S3 / EMR / Redshift
- AWS Lambda
- Azure Synapse
- Snowflake
- BigQuery
- Apache Airflow
- Vertex AI
- Step Functions
MLOps & Viz
- Docker
- Kubernetes
- Jenkins
- CI/CD
- MLflow
- Git
- Tableau
- Power BI
Credentials & focus
Formal education, certifications, and the domains where the work is happening.
Education
M.S.
Computer Science
SUNY Albany · Albany, NY
Graduate coursework in machine learning, data systems, and applied AI.
B.Tech.
Computer Science
Bharath University · India
Domains
- Financial MLRisk prediction, anomaly & fraud detection
- GenAIRAG, agentic workflows, prompt engineering
- MLOpsCI/CD for ML, MLflow, Vertex AI, monitoring
- BackendServerless, AWS Lambda, Step Functions, REST
- DataPySpark, Kafka, Snowflake, vector stores
- AnalyticsTableau, Power BI, EDA, hypothesis testing
Certifications · tap a card to open the PDF
Let's build
something that matters.
Open to senior AI/ML and Generative-AI roles. Happy to consult on financial ML, RAG architecture, or productionizing LLM systems.