Long before Deepti Yenireddy asked a machine to interpret a construction drawing, she learned how unforgiving physical data can be. At Schlumberger, fresh from an electrical engineering degree at IIT Madras, she built and ran hardware tools that gathered information in oil fields. At Shell, she collected and interpreted readings across countries. The instruments had to survive the world as it was: remote, expensive, pressured and indifferent to a clean software demo.
That early work is the quiet prologue to Boon, the enterprise AI company Yenireddy now leads. Boon began by automating fleet and logistics operations. Its public product today is aimed at construction, where estimators face an equally physical kind of complexity: drawings layered across pages, specifications that qualify what a symbol means, addenda that land near a deadline, and decisions whose cost arrives in steel, wire, concrete and labor.
The markets look different. The founder's habit is familiar. Find an expert surrounded by scattered information. Learn the workflow closely enough to speak its language. Let software do the repetitive reading, counting and cross-referencing. Then return the result with enough evidence that a person can check it.
“Nobody trusts a black box” that can make decisions for them without showing the different aspects.Deepti Yenireddy, on the adoption problem in AI
The girl beside the tennis court
Yenireddy grew up in a middle-class family in Hyderabad. Her father coached tennis; her mother was a homemaker. She played, and would eventually win a gold medal for IIT Madras, but her own telling includes a more revealing detail: at the stadium, she was drawn to the library. Physics held her. A top-ten finish in a high-school mathematics Olympiad helped settle the choice of science and engineering over medicine.
In 2002 she entered IIT Madras to study electrical engineering, one of six women in a class of 120. The place made building feel available. In her first year she soldered rather than merely reading about circuits. She fixed a tool to a bicycle to measure distance. She explored desalination ideas and imagined starting a company. The scale was modest; the conversion of theory into a working object was the point.
Field engineering followed. At Schlumberger she worked with hardware that collected data for oil and gas customers. At Shell she interpreted data from different tools and traveled to Bahrain, Egypt and Scotland. She was often the youngest person and the only woman on teams operating in remote, high-pressure settings. She has recalled needing to work harder to be taken seriously. The experience gave her a founder's tolerance for hearing no, but it also left a sharper question: what does it take for a person to trust a system, a colleague or a result?
A founder learns to show the work
In 2009, Yenireddy moved to the United States and entered the Cox School of Business at Southern Methodist University. After earning an MBA in finance in 2011, she worked around early-stage investing, including roles connected to Oppenheimer Funds and USGT Investors. She saw how companies were formed and financed. In 2015, she stepped from evaluating startups into building one.
My Ally began with a daily irritation: scheduling meetings. Yenireddy and a friend from IIT Madras built an AI executive assistant using natural-language processing and machine learning. The company later narrowed into recruiting, automating conversations and interview coordination for employers. Its clients included Booking.com and SAP, and its investors included Storm Ventures and Gokul Rajaram.
The consequential product choice was not merely the assistant's ability to act. It was the decision to expose its reasoning. My Ally built an activity center where a user could see the parameters behind the assistant's decisions and change them. Accuracy mattered, but visibility and control made accuracy usable. When Phenom acquired My Ally in September 2020, Yenireddy continued the work as vice president of product for conversational AI, now at a larger scale.
That principle followed her to Samsara, where she became a senior director of product and led work including telematics and international products. The role put her close to fleets and other businesses whose software sits on top of moving vehicles, field workers and thin operating margins. Customers did not want another isolated application. They wanted fewer seams between routing, fuel, compliance, orders and the back office.
Boon and the second-employee idea
Yenireddy founded Boon in 2023. Her early description was concrete: think of its AI agent as a second employee in the back office, doing critical work while people focus on tasks that make the business money. The company connected data across a fleet's fragmented applications and attacked workflows such as order entry, fuel planning and compliance review.
The response was enough to finance the next chapter. In December 2024, Boon announced $20.5 million across a previously undisclosed $5 million seed and a $15.5 million Series A led by Marathon and Redpoint. At that point the company said paying customers represented 35,000 drivers and 10,000 vehicles, and that Boon had reached a $1 million annual revenue run rate after nine months.
Boon's current construction focus brings Yenireddy's career into a revealing loop. A plan set is not just a pile of PDFs. Symbols vary by architect. Building codes vary by place and project. A line may continue across pages. A specification can reverse the apparent meaning of a drawing. The problem involves language, vision, geometry and trade knowledge at once.
This is why Yenireddy argues for specialized systems rather than a general chatbot pointed at a blueprint. Boon generates synthetic, code-compliant plan data to supplement limited real-world training material. Its products perform takeoffs, identify conflicts, track bids and compare information across documents. The output is designed for review: an estimator should be able to see where an answer came from, correct it and keep control.
“The key thing to do is try a few specialized systems and build trust in them.”Deepti Yenireddy, speaking to construction estimators
The expert moves up the stack
The human consequence matters. Yenireddy does not frame estimation as a job on the edge of extinction. Her argument is that measurement and cross-referencing can move to software while risk, engineering judgment and project strategy stay with people. One Boon customer described in a podcast worked through a project from his couch on a Friday evening, directing an agent to review specifications, flag conflicts and measure feeders. The anecdote is less about a magical coworker than a changed division of labor: the estimator directs and reviews instead of manually tracing every step.
Start with the expert's repeated task, not the model's newest capability. Put the evidence beside every answer. Let corrections travel back into the system. Measure the result in capacity returned to the team.
That stance is consistent with Yenireddy's older writing and interviews about recruiting AI. She has argued that people need to see decisions, retain control and understand the return on investment. The object being read has changed from a candidate conversation to a set of construction documents. The social contract has not.
Her operating style is equally direct. A podcast profile reported that she spends roughly 90 percent of her time with customers and pushes iterations out in days, not months. She has also described a three-priority rule: decide the few things that form the North Star, then abandon work that does not align. Customer proximity supplies the raw signal; hard prioritization decides what becomes product.
There is a personal symmetry here. The student who preferred the stadium library still reads physics and discusses problems with her son. The engineer who once attached a measuring tool to a bicycle now builds systems that measure walls, beams and feeders. The founder who exposed an AI recruiter's logic now asks construction AI to cite its work. Each chapter gets more ambitious, but none escapes the discipline of measurement.
Yenireddy's aspiration for Boon is expressed in operating terms: more capacity, wider margins and less time lost to essential but repetitive work. The important word is not autonomous. It is accountable. In the physical world, the final answer eventually becomes a route, a hire, a bid or a building. Someone still has to stand behind it.