Akshita Bhardwaj

AI/ML • Junior • Management: 1-2 years • Greater Noida, India • UTC+05:30 • Relocation • Work permits: India • Can start now

AI Engineer

5+ years exp

5+ years ML exp

Management: 1-2 years

Python

SQL

Data Pipeline & Feature Engineering: 4/10

Experimentation & Evaluation: 4/10

Active 7 days ago

Invite to interview

Download CVCV

Overview

Technical skills

Projects

Timeline

Roles

Overview

LLM engineer specializing in retrieval-augmented systems and agentic pipelines at a solid mid-level with an emphasis on practical RAG tooling and demo-grade evaluation. The strongest proven skill is building end-to-end RAG and agent demos with mitigation and evaluation, demonstrated by the RAGEvaluator and HallucinationMitigator implementations and the agent graph with confidence gating. There is limited evidence of custom model training, large-scale MLOps, benchmarking at scale, or low-level efficiency engineering in the public code.

Phone

Technical skills

Languages

Python

SQL

AI/ML

Scikit-learn

Claude

Gemini

GPT-4

LLM

RAG

NumPy

Pandas

Transformers

Embeddings

LangChain

LangGraph

Streamlit

Gradio

TensorFlow

XGBoost

Hugging Face

OpenAI

Keras

Anthropic

SciPy

Databases

PostgreSQL

FAISS

MySQL

Analytics

Matplotlib

Power BI

Plotly

Seaborn

Other

Requests

Beautiful Soup

GitHub

Tableau

Docker

Rest API

Git

AI Agents

Machine Learning

Agentic Workflows

Function Calling

Semantic Search

ETL/ELT

Time Series

Projects

EEG Signal Analysis and Statistical Modelling

Performed signal processing and statistical analysis on EEG data as an MSc dissertation project. Applied FIR filtering and frequency analysis using FFT and Welch power spectral density estimation. Used statistical tests such as Shapiro-Wilk, Kruskal-Wallis, and Kendall Tau correlation to examine patterns in the signals.

Supervised ML Models

Implemented supervised learning models for classification and regression. Included feature selection, cross-validation, and hyperparameter tuning to compare model performance. Focused on producing measurable results suitable for reporting and iteration.

Supply Chain Performance Analytics

Created an analytical pipeline for supply-chain datasets to identify performance patterns, exceptions, and potential loss indicators. Combined data processing with KPI-style reporting to help operational teams review trends and anomalies. Delivered the workflow with API-accessible components for use in downstream tasks.

Agentic Query Pipeline

Developed a multi-step question workflow that decomposes queries, uses tools, and synthesizes grounded answers from retrieved context. Implemented confidence gating and fallback behavior to reduce low-confidence outputs. Exposed the workflow through REST API interfaces for easier integration.

RAG-Based Insight Platform

Built a document ingestion and retrieval workflow with chunking, metadata tagging, and vector indexing to enable search over source material. Added source attribution with claim extraction and provenance scoring. Included evaluation for retrieval accuracy, response consistency, and latency, packaged with a reproducible Docker setup.

Timeline

Founder and AI Systems Lead • Head+

Duovarsity • Cofounder

Jan 2025 to Apr 2026 1 Year 3 Months In office

Led development of a retrieval-augmented generation (RAG) pipeline using LLMs and embedding-based retrieval. Built citation-linked response workflows with provenance scoring to improve source traceability. Implemented multi-step agentic workflows for query refinement and tool-assisted response generation, plus evaluation routines to measure retrieval quality, response consistency, and latency.

RAG

Embeddings

GPT-4

Gemini

Claude

Function Calling

University of Aberdeen

Master's Degree • Data Science

2023–2024 Aberdeen, Scotland

Data Scientist • Middle

Delhivery • Full-Time

Dec 2021 to Dec 2022 1 Year Delhi In office

Automated more than 80 recurring operational reports using Python to reduce manual work for logistics teams. Wrote and maintained SQL queries to extract, combine, and analyze warehouse and delivery data for performance monitoring. Developed operational dashboards and KPI reporting, including risk-focused views for fraud and exception tracking.

Python

SQL

Data Scientist Intern • Junior

Delhivery • Internship

Oct 2021 to Dec 2021 2 Months Delhi In office

Performed exploratory analysis and data validation for logistics and warehouse datasets supporting recurring reporting. Built SQL queries and reporting scripts to support analytical outputs. Assisted in preparing findings and communicating results to operational stakeholders.

Python

SQL

Blockchain Intern • Junior

CFTE • Internship

Aug 2021 to Sep 2021 1 Month Partially remote

Researched blockchain and fintech concepts including DeFi protocols, automated market makers, and distributed ledger fundamentals. Reviewed transaction and token-economics topics and produced structured summaries for learning and research purposes.

Data Science Intern, GRIP • Junior

The Sparks Foundation • Internship

Jun 2021 to Jul 2021 1 Month Delhi In office

Implemented supervised classification and regression exercises using scikit-learn with attention to feature preparation and model validation. Documented model approaches and results in a clear format for non-technical readers. Supported experimentation through cross-validation and performance comparison.

Python

Scikit-learn

Guru Gobind Singh Indraprastha University (GGSIPU)

Bachelor's Degree • Computer Science Engineering

2016–2020 New Delhi, India

Junior AI/ML Engineer Confidence: Medium LLM Engineer

Model Architecture & Training

2/10

How well models are designed and trained

Minimal model architecture or training work is present; the code relies on pretrained embedding models and inference rather than custom training loops or novel architectures.

Evidence

rag-insight-platform/src/vector_store.py: VectorStore._embed uses SentenceTransformer for embeddings

rag-insight-platform/src/mitigation.py: HallucinationMitigator loads SentenceTransformer and uses util.cos_sim for scoring

Data Pipeline & Feature Engineering

4/10

How data is prepared for models

Clear, practical data ingestion and chunking pipeline with metadata propagation and sensible chunking heuristics; this is a solid mid-level data pipeline implementation for RAG use cases.

Evidence

rag-insight-platform/src/ingestion.py: ingest_text, ingest_file, ingest_url functions with metadata and source handling

rag-insight-platform/src/chunking.py: chunk_text and chunk_documents implement sentence splitting, overlap, and chunk metadata

Experimentation & Evaluation

4/10

How results are measured and tested

Honest, reproducible evaluation scaffolding with per-query metrics, benchmark loading, and a runnable run_benchmark implementation suitable for lightweight experiments and demos.

Evidence

rag-insight-platform/src/evaluation.py: RAGEvaluator.evaluate_query and run_benchmark compute latency, similarity and faithfulness metrics

rag-insight-platform/evals/sample_evaluation.json: contains benchmark queries used by the evaluation runner

MLOps & Deployment

3/10

How models are shipped to production

Basic serving and deployment patterns are present - Streamlit and Gradio demos and index save/load are implemented - but there is limited evidence of production MLOps, CI/CD, monitoring, or model lifecycle automation.

Evidence

rag-insight-platform/app.py: Streamlit app wiring that runs evaluations and queries against the pipeline

agentic-query-pipeline/app.py: Gradio demo and streaming trace panel

rag-insight-platform/src/vector_store.py: save and load methods for persisted FAISS index

Computational Efficiency

2/10

How efficiently computing resources are used

Some attention to efficiency through use of FAISS and compact embedding models, but no evidence of systematic profiling, GPU optimization, batching strategies, quantization, or memory/GPU engineering.

Evidence

rag-insight-platform/requirements.txt: faiss-cpu listed as dependency

rag-insight-platform/src/vector_store.py: faiss index usage in search/build

Research Depth & Innovation

3/10

Depth of research and new ideas

Shows applied engineering around mitigation and confidence gating and implements provenance scoring and an agent graph, but lacks research-grade novelty or reproduced SOTA experiments.

Evidence

rag-insight-platform/src/mitigation.py: HallucinationMitigator with claim extraction and provenance scoring

agentic-query-pipeline/src/confidence.py: evaluate_confidence implements a multi-factor confidence gate

agentic-query-pipeline/src/graph.py: build_query_graph and CompiledGraph implement the agent execution model and streaming

Expertise

AI Agents & Agentic Workflows• Junior

RAG• Junior

Industries

Artificial Intelligence• Junior

Software• Junior

Technologies

SQL• since 2021

MySQL

PostgreSQL

LangGraph

LangChain

Claude• since 2025

FAISS

XGBoost• since 2026

Embeddings• since 2025

Scikit-learn• since 2021

SciPy

Function Calling

AI Agents

Gradio

Transformers

TensorFlow• since 2026

Pandas

NumPy

Keras• since 2026

Git• since 2026

Docker

Gemini• since 2025

LLM

RAG• since 2025

Streamlit

Time Series

OpenAI

Anthropic

Hugging Face

GitHub• since 2026

GPT-4• since 2025

Semantic Search

Agentic Workflows

Machine Learning

Recommendations

Develop production-ready RAG pipelines and evaluation consoles that include provenance scoring and claim-level analyses.
Implement agentic workflows and confidence-gated tool orchestration for multi-step queries and streaming traces.
Extend the current demos into production MLOps by adding CI/CD, index versioning, monitoring, and performance profiling for vector search and embedding pipelines.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

rag-insight-platform agentic-query-pipeline 230akshita

Intern Data Scientist Confidence: Low Generalist

An early-career generalist focused on Python-based analytics and demo retrieval systems. The most visible artifacts are demo analytics and RAG scaffolds, but file ownership and boilerplate signals indicate these are largely template or vendor-supplied so no original engineering depth is proven. There is insufficient human-authored code, tests, or CI to evaluate system design, production readiness, or rigorous data-science practices.

Statistical Rigor

Correct use of statistics

Not evidenced in public code

Data Wrangling & Cleaning

Preparing and cleaning data

Not evidenced in public code

Exploratory Analysis & Visualization

Exploring and visualizing data

Not evidenced in public code

Predictive Modeling

Building models that predict

Not evidenced in public code

Business Insight & Impact

Turning analysis into business value

Not evidenced in public code

Reproducibility & Notebook Hygiene

Clean, repeatable analysis

Not evidenced in public code

Recommendations

Assign to prototype or extend Streamlit dashboards and craft small data exploration notebooks to validate UX and metrics.
Have the developer implement ingestion/ETL tasks and connectors (CSV/URL/file ingestion) with unit tests and clear error handling.
Task with adding reproducibility: pin environment, add tests, CI, and data versioning for a single pipeline to demonstrate ownership.
Use for supervised, mentored work on vector store integration (FAISS) and simple evaluation harnesses under senior guidance.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

rag-insight-platform 230akshita supply-chain-analytics

Intern Backend Developer Confidence: Low Generalist

Junior Python engineer focused on building retrieval-augmented generation and agentic pipelines with demo frontends. The clearest proven skill is practical Python tooling and web-demo wiring as shown by requirements.txt and the agentic query graph and demo wiring (for example src/graph.py and app.py). There is little to no evidence of production-grade backend engineering concerns such as migration history, transactional boundaries, robust concurrency controls, or end-to-end API/service contracts in human-authored files.

API Design

How well APIs are designed

Not evidenced in public code

Data Layer & Database

Working with databases

Not evidenced in public code

Scalability & Performance

Handling load and speed

Not evidenced in public code

System Architecture

Overall system structure

Not evidenced in public code

Security & Auth

Protecting data and access

Not evidenced in public code

Reliability & Observability

Stability and monitoring

Not evidenced in public code

Expertise

Python• Intern

Industries

Artificial Intelligence• Junior

Technologies

Python• since 2021 • Junior

Rest API

Beautiful Soup

Requests

Recommendations

Prototype RAG and agent demos, ingestion pipelines, and evaluation harnesses where quick iteration and user-facing demos are primary goals.
Build proof-of-concept agent flows and tool orchestrations that integrate third-party REST tools and mock services.
Work on ingestion and chunking pipelines for text corpus preparation and simple vector store experiments.

Repositories

The developer's experience in this domain has been verified based on AI analysis of the following repositories:

rag-insight-platform agentic-query-pipeline 230akshita