I do not trust AI output until it is evaluated.
I am building toward systems where generated code, extracted data, and model responses are tested, scored, and validated before humans rely on them.
Akshit Mittal
Software Engineer
Currently
At BlackRock, building data-heavy internal tools and AI extraction workflows. On the side, thinking out loud about software engineering and AI on X.
Overview
Software Development Engineer @BlackRock
B.Tech EE + CS Minor @Delhi Technological University
Social Links
About
I am a final-year B.Tech Electrical Engineering student with a CS minor at Delhi Technological University, Class of 2026.
I am currently a Software Development Engineer at BlackRock in Gurugram, working around full-stack development, ETL pipelines, OCR/NLP-based document extraction, and agentic AI workflows.
Outside work and college, I am building a sharper engineering practice around AI, systems, and practical developer tooling, and writing about software and AI on X.
GitHub Contributions
Stack
- TypeScript
- JavaScript
- Python
- C++
- C
- Java
- SQL
- HTML/CSS
- OpenAI API
- Gemini API
- LangChain
- Vector Search
- AI Evals
- OCR/NLP
- spaCy
- Pandas
- pytesseract
- pdfplumber
- React.js
- Next.js
- Node.js
- Express.js
- FastAPI
- Flask
- Tailwind CSS
- Material UI
- Prisma
- Convex
- REST APIs
- Webhooks
- PostgreSQL
- MySQL
- MongoDB
- Snowflake
- AWS
- EC2
- S3
- CloudFront
- Docker
- Vercel
- Git
- GitHub
- Linux
- Postman
- Figma
- Cursor
- JetBrains IDEs
- LeetCode Knight · 700+
Experience
BlackRock
- Engineered a full-stack Fideuram Dashboard with React, TypeScript, FastAPI, and Pandas to consolidate quarterly holdings from Airtable with daily Aladdin ATP trade feeds, enabling real-time exposure and allocated-capital computation across 500+ portfolio positions and reducing analyst reconciliation time by about 60%.
- Built an automated ETL pipeline with Python, requests, airtable-python-wrapper, and openpyxl to ingest and normalize valuation reports into Airtable, eliminating manual data entry and reducing update latency from days to minutes.
- Developed an OCR and NLP KPI extraction pipeline using pytesseract, pdfplumber, spaCy, and regex to parse Quarterly Portfolio Reviews and feed structured outputs into ASIMOV, BlackRock's internal agentic AI platform, improving portfolio manager insight retrieval speed by 45%.
- Designed Pandas-based trade-to-holding reconciliation logic to map post-quarter transactions onto last known positions and produce exposure deltas and PnL attribution metrics for downstream agent workflows and dashboard visualizations.
- React
- TypeScript
- FastAPI
- Pandas
- ETL
- OCR
- NLP
- Agentic AI
- Airtable
- Aladdin ATP
Tech Mahindra
- FastAPI
- React
- LangChain
- Gemini API
- MongoDB
- Vector Search
- Whisper
- Docker
- Material UI
Defence Research and Development Organisation (DRDO)
- Embedded C
- RM42L432BPZT
- State Machines
- ADC
- Sensor Calibration
- Fritzing
- Hardware Prototyping
Education
Delhi Technological University
- Degree
- B.Tech Electrical Engineering, CS Minor
- Period
- 2022-2026
- Location
- Delhi, India
Final-year undergraduate combining an electrical engineering base with a computer science minor and software engineering practice. CGPA: 9.20/10.
- Data Structures & Algorithms
- Object-Oriented Programming
- Discrete Structures
- Machine Learning
- Cloud Computing
- Computer Architecture
- Programming Fundamentals
- Numerical Optimization
- Mathematics
- Digital Circuits
- Microprocessors & Microcontrollers
- Control Systems
- Network Analysis
- Power Systems
- Power Electronics
Projects(3)
- Built an agentic code trust platform that turns prompt, code, and GitHub inputs into an automated generate, red-team, execute, evaluate, and repair loop for stress-testing AI-generated code across correctness, robustness, security, performance, and code quality.
- Architected the Convex backend to orchestrate multi-stage AI workflows with persistent run state, version history, adversarial attack cases, evaluation evidence, fix suggestions, live queries, scheduled actions, and GitHub webhook-triggered re-analysis of changed JavaScript and TypeScript files.
- Designed a real-time trust cockpit with live stage feeds, iteration telemetry, score deltas, failure evidence, execution-mode tracking, and best-version promotion, giving users transparent visibility into how AI code improves across repair cycles.
- React
- TypeScript
- Convex
- OpenAI API
- Vite
- Agentic AI
- GitHub Webhooks
- Code Evaluation
Not Testimonials. Signals.
Early-career engineer, so I am not borrowing credibility. I am showing receipts: projects, research, systems work, and the engineering habits I am building.
I build AI that survives messy data.
At BlackRock, I worked on validation-first pipelines for reports, holdings, trades, and extracted portfolio signals instead of assuming clean inputs.
